Search

Cropland non-agriculturalization (CNA) refers to the conversion of cropland into construction land, woodland/garden/grassland, water body, or other non-agricultural land, which ultimately disrupts local agroecosystems and the cultivation and production of crops. Remote sensing technology is an important tool for large-area CNA detection, and remote sensing based methods that can be used for this task include the time-series analysis method and change detection from bi-temporal images. In particular, change detection methods using high-resolution remote sensing imagery have great potential for CNA detection, but enormous challenges do still remain. The large intra-class variance of cropland with different phenological stages and planting patterns leads to cropland areas being difficult to identify effectively, while certain features can be misidentified because they are similar to cropland, resulting in false alarms and missed detections in the results. There is also a lack of large-scale CNA datasets covering multiple change scenarios as data support. To address these problems, a lightweight model focused on CNA detection (CNANet) is proposed in this paper. Specifically, the uniquely crafted represent-consist-enhance (RCE) module is seamlessly integrated between the encoder and decoder components of CNANet to perform a contrast operation on the deep features extracted by the feature extractor. The RCE module is specifically designed to aggregate multiple cropland representations and extend the cropland representations from the confusing background, to achieve the purpose of reducing the intra-class reflectance differences and enhancing the model's perception of cropland. In addition, a large-scale high-resolution cropland non-agriculturalization (Hi-CNA) dataset was built for the CNA identification task, with a total of 6797 pairs of 512 × 512 images with semantic annotations. Compared to the existing datasets, the Hi-CNA dataset has the advantages of multiple phenological stages, multiple change scenarios, and multiple annotation types, in addition to the large data volume. The experimental results obtained in this study show that the benchmark methods tested on the Hi-CNA dataset can all achieve a good accuracy, proving the high-quality annotation of the dataset. The overall accuracy and F1-score of CNANet with the default settings reach 93.81 % and 78.9 %, respectively, achieving a superior accuracy, compared to the other benchmark methods, and demonstrating stronger perception of cropland changes. In addition, in two selected verification regions within the large-scale real-world CNA mapping results, the F1-score is 83.61 % and 50.87 %. The Hi-CNA can be downloaded from http://rsidea.whu.edu.cn/Hi-CNA_dataset.htm. [ABSTRACT FROM AUTHOR]

Urban development in South America has experienced significant growth and transformation over the past few decades. South America's urban development and trees are closely interconnected, and tree cover within cities plays a vital role in shaping sustainable and resilient urban landscapes. However, knowledge of urban tree canopy (UTC) coverage in the South American continent remains limited. In this study, we used high-resolution satellite images and developed a semi-supervised deep learning method to create UTC data for 888 South American cities. The proposed semi-supervised method can leverage both labeled and unlabeled data during training. By incorporating labeled data for guidance and utilizing unlabeled data to explore underlying patterns, the algorithm enhances model robustness and generalization for urban tree canopy detection across South America, with an average overall accuracy of 94.88% for the tested cities. Based on the created UTC products, we successfully assessed the UTC coverage for each city. Statistical results showed that the UTC coverage in South America is between 0.76% and 69.53%, and the average UTC coverage is approximately 19.99%. Among the 888 cities, only 357 cities that accommodate approximately 48.25% of the total population have UTC coverage greater than 20%, while the remaining 531 cities that accommodate approximately 51.75% of the total population have UTC coverage less than 20%. Natural factors (climatic and geographical) play a very important role in determining UTC coverage, followed by human activity factors (economy and urbanization level). We expect that the findings of this study and the created UTC dataset will help formulate policies and strategies to promote sustainable urban forestry, thus further improving the quality of life of residents in South America. [ABSTRACT FROM AUTHOR]

Semantic understanding of high-resolution remote sensing (RS) images is of great value in Earth observation, however, it heavily depends on numerous pixel-wise manually-labeled data, which is laborious and thereby limits its practical application. Semi-supervised semantic segmentation (SSS) of RS images would be a promising solution, which uses both limited labeled data and dominant unlabeled data to train segmentation models, significantly mitigating the annotation burden. The current mainstream methods of remote sensing semi-supervised semantic segmentation (RS-SSS) utilize the hard or soft pseudo-labels of unlabeled data for model training and achieve excellent performance. Nevertheless, their performance is bottlenecked because of two inherent problems: irreversible wrong pseudo-labels and long-tailed distribution among classes. Aiming at them, we propose a decoupled weighting learning (DWL) framework for RS-SSS in this study, which consists of two novel modules, decoupled learning and ranking weighting, corresponding to the above two problems, respectively. During training, the decoupled learning module separates the predictions of the labeled and unlabeled data to decrease the negative impact of the self-training of the wrongly pseudo-labeled unlabeled data on the supervised training of the labeled data. Furthermore, the ranking weighting module tries to adaptively weight each pseudo-label of the unlabeled data according to its relative confidence ranking in its pseudo-class to alleviate model bias to majority classes as a result of the long-tailed distribution. To verify the effectiveness of the proposed DWL framework, extensive experiments are conducted on three widely-used RS semantic segmentation datasets in the semi-supervised setting. The experimental results demonstrate the superiority of our method to some state-of-the-art SSS methods. Our code will be available at https://github.com/zhu-xlab/RS-DWL. [ABSTRACT FROM AUTHOR]

Extracting building footprints from extensive very-high spatial resolution (VHSR) remote sensing data is crucial for diverse applications, including surveying, urban studies, population estimation, identification of informal settlements, and disaster management. Although convolutional neural networks (CNNs) are commonly utilized for this purpose, their effectiveness is constrained by limitations in capturing long-range relationships and contextual details due to the localized nature of convolution operations. This study introduces the masked-attention mask transformer (Mask2Former), based on the Swin Transformer, for building footprint extraction from large-scale satellite imagery. To enhance the capture of large-scale semantic information and extract multiscale features, a hierarchical vision transformer with shifted windows (Swin Transformer) serves as the backbone network. An extensive analysis compares the efficiency and generalizability of Mask2Former with four CNN models (PSPNet, DeepLabV3+, UpperNet-ConvNext, and SegNeXt) and two transformer-based models (UpperNet-Swin and SegFormer) featuring different complexities. Results reveal superior performance of transformer-based models over CNN-based counterparts, showcasing exceptional generalization across diverse testing areas with varying building structures, heights, and sizes. Specifically, Mask2Former with the Swin transformer backbone achieves a mean intersection over union between 88% and 93%, along with a mean F-score (mF-score) ranging from 91% to 96.35% across various urban landscapes. [ABSTRACT FROM AUTHOR]

Machine learning classifiers have been widely used for crop mapping; however, employing numerous features can reduce classifier accuracy. Feature selection is a widely recognized technique that addresses the issue of high dimensionality by choosing the most relevant subset of features with the highest degree of relevance and practicality. Therefore, the optimal combination of feature selection methods and classifiers for high-precision crop mapping remains an open problem. In this study, we introduce a novel hybrid feature selection approach to obtain the optimal subset of features for effective crop mapping. Initially, spectro-temporal remote sensing features from the dataset were ranked using two distinct feature selection techniques: Mutual Information (MI) and ReliefF filter. The most relevant features from each filter-based approach were combined into a union subset. Subsequently, the Gray Wolf Optimization (GWO), a metaheuristic optimization technique, was applied to enhance the feature set generated in the initial step. In the final stage, a random forest classifier leveraged the optimized feature subset for accurate crop type prediction. The effectiveness of the proposed approach was evaluated in Behiera governorate, Egypt. Comparative analysis against existing crop mapping methods demonstrated the superior performance of the proposed approach, achieving an accuracy rate of 82%. Furthermore, this approach showcased a significant reduction in the feature count, streamlining the set from 308 down to only 26 features. [ABSTRACT FROM AUTHOR]

• Proposed a Hybrid Model for Paddy Crop prediction based on phenological parameters. • CNN-LSTM attained RMSE value of 0.219 t/ha compared to other existing methods. • Random Forest method achieved highest accuracy of 96.6% in paddy classification. Rice is one of the predominant food sources to fulfill the dietary requirements of well-being in India. Therefore, accurate and timely paddy crop yield prediction is crucial to ensure the food security of the country. In this direction, the present study proposed a hybrid deep-learning method based on Conv-1D and LSTM layers using the classification-derived phenological with meteorological parameters for paddy crop yield prediction. The paddy crop classification has been conducted using high-resolution (10 m) multispectral imagery based on GPS coordinates collected during the paddy field visits to extract the phenological parameters for input to the prediction model. In this context, the efficiency of Random Forest, Naïve Bayes, SVM, and Gradient Tree boost classifiers was assessed. Furthermore, we have also analyzed the accuracy of Landsat-8, Sentinel-1 GRD, and Sentinel-2 satellite imagery in paddy crop classification based on area estimation. The Statistical Abstract of Haryana was utilized to validate the paddy crop area estimation and yield prediction. The classification outcomes showed that the Random Forest method attained the highest accuracy of 96.6 % compared to other GEE-based classifiers. The proposed Hybrid Deep learning approach achieved an RMSE value of 0.219 t/ha compared to CNN, LSTM, CNN-Bi-LSTM, and Regression techniques for crop yield prediction. The study conclusion highlighted that the sentinel-2 satellite imagery performed well and found that the proposed hybrid approach provided an alternative for paddy crop yield prediction. [ABSTRACT FROM AUTHOR]

The tradeoff among spatial, temporal, and spectral resolution of remote sensing (RS) images due to sensor properties limits the development of RS applications. Most image enhancement studies tend to focus on either spatio-temporal fusion or spatio-spectral fusion. As a more comprehensive solution, spatial–temporal-spectral fusion (STSF) is complicated but its potential is worth to be further explored. In this study, we propose a novel STSF method from the perspective of multi-temporal Pansharpening. Canny edge extraction is applied to Panchromatic (PAN) images to identify edges while avoiding the disruption of multi-temporal land cover changes. We then build a TemPanSharpening net (TPSnet) which only uses one high-spatio-low-spectra-temporal PAN and one low-spatio-high-spectra-temporal multispectral image as input. TPSnet follows a super-resolution structure and embeds two basic modules: residual-in-residual dense blocks (RRDB) and convolutional block attention module (CBAM). A series of interior ablation experiments were conducted on TPSnet and we also compared it with some representative spatio-temporal fusion, Pansharpening, and STSF algorithms. TPSnet presented satisfactory performance on complicated meter-level ground surfaces according to the quantitative evaluation result, and it demonstrated excellent robustness to land cover change. [ABSTRACT FROM AUTHOR]

• ClearSCD considers the bidirectional dependency between semantics and change. • A BSCC mechanism establishes the correlation of bi-temporal semantic features. • A deep CVAPS module enhances the efficiency of change detection. • A SACL module optimizes feature space of semantics explicitly. • A new semantic change detection dataset, LsSCD, was proposed by this study. The Earth has been undergoing continuous anthropogenic and natural change. High spatial resolution (HSR) remote sensing imagery provides a unique opportunity to accurately reveal these changes on a planetary scale. Semantic change detection (SCD) with HSR imagery has become a common technique for tracking the evolution of land surface types at a semantic level. However, existing SCD methods rarely model the dependency between semantics and changes, resulting in suboptimal accuracy in detecting complicated surface changes. To address this limitation, we propose ClearSCD, a multi-task learning model that leverages the mutual gain relationship between semantics and change through three innovative modules. The first module interprets semantic features at different times into posterior probabilities for surface types to detect binary change information; the second module learns the correlation between surface types over time and the binary change information; a semantic augmented contrastive learning module is used as the third module to improve the performance of the other two modules. We tested ClearSCD's performance against state-of-the-art methods on benchmark datasets and a real-world scenario (named LsSCD dataset), showing that ClearSCD outperformed the alternatives on mIoU sc metrics by 1.23% to 19.34%. Furthermore, ablation experiments demonstrated the unique contribution of the three innovative modules to performance improvement. The high computational efficiency and robust performance over diverse landscapes demonstrate that ClearSCD is an operational tool for detecting detailed land surface changes from HSR imagery. Code and LsSCD dataset available at https://github.com/tangkai-RS/ClearSCD. [ABSTRACT FROM AUTHOR]

Excessive discharges of nitrogen and phosphorus nutrients lead to eutrophication in coastal waters. Optical remote sensing retrieval of the key eutrophication indicators, namely dissolved inorganic nitrogen concentration (DIN), soluble reactive phosphate concentration (SRP), and chemical oxygen demand (COD), remains challenging due to lack of distinct spectral features. Although machine learning (ML) has shown the potential, the retrieval accuracy is limited, and the interpretability is insufficient in terms of the black-box characteristics. To address these limitations, based on robust and explainable ML algorithms, we constructed models for retrieving DIN, SRP, and COD over coastal waters of Northern South China Sea (NSCS), which is experiencing prominent eutrophication. Retrieval models based on classification and regression trees (CART) ML algorithms were developed using 4038 groups of in situ observations and quasi-synchronous satellite images. A comparison of CART algorithms, including Random Forest, Gradient Boosting Decision Tree, and eXtreme Gradient Boosting (XGBoost), indicated the highest retrieval accuracy of XGBoost for DIN (R2 = 0.88, MRE = 24.39 %), SRP (R2 = 0.92, MRE = 33.27 %), and COD (R2 = 0.75, MRE = 18.58 %) for validation dataset. On the basis of spectral remote sensing reflectance, further inputs of ocean physio-chemical properties, spatio-temporal information, and inherent optical properties may reduce retrieval errors by 30.16 %, 19.85 %, and 3.95 %, respectively, and their combined use reduced errors by 54.71 %. Besides, explainable ML analysis characterized the contribution of input features and enhanced the transparency of ML black-box models. Based on the proposed models, 27,278 satellite images and spatio-temporal reconstruction method, 1-km resolution gap-free daily DIN, SRP, and COD products were constructed from 2002 to 2022 for the coastal waters of NSCS. Under the influence of urbanization and river discharge, nitrogen and phosphorus concentrations in this area were found to have increased by 6.09 % and 11.04 %, respectively, over the past 21 years, with the fastest rise in the Pearl River Estuary, where the eutrophic water area had shown an increase rate of approximately 112.66 km2/yr. The proposed robust and explainable ML retrieval models may support ocean environment management and water quality monitoring by providing key eutrophication indicators products over coastal waters. [ABSTRACT FROM AUTHOR]

Forest phenology, as a sensitive indicator of a forest's response to climate change and variability, has long been monitored using remote sensing, yet has seldom been interpreted or validated with spatially compatible, community-level field phenological observations. In temperate deciduous forests, multiple spring phenological phases are critical for modeling carbon storage and biogeochemical cycles. The simultaneous detection of all these critical phenological phases at the community level remains a long-standing challenge. To tackle these challenges, the objective of this study is to develop a novel satellite-field phenological bridging framework for characterizing all key spring phenological phases of a fragmented deciduous forest using multi-scale satellite time series. The framework consists of four key components: deep learning-based spatiotemporal image fusion, satellite-based forest phenology modeling, satellite-based forest phenological metric extraction, and field-based forest community phenological characterization. With the devised framework, we extract a total of 24 satellite phenological metrics of Trelease Woods, a forest fragment near Urbana, IL, USA. From weekly field phenological observations of 148 canopy trees of 15 common species of the forest community, we devise three summative field phenological indices to characterize community-level phenological states in spring. Under the satellite-field bridging framework, events during each key spring forest phenological phase (i.e., bud swell, budburst/leafing out, leaf expansion, and leaf maturity) are successfully detected using the fusion imagery (MAE from 1.1 to 2.9 days and bias from -2.4 to 1.5 days). However, the satellite detection of the earliest field events may be influenced by understory plants, soil background, and snow. The subsequent multi-scale, satellite phenological analysis underscores the importance of taking into account spatial scale and representation from both satellite and field phenological perspectives in building corresponding bridging relationships. Among the extracted pheno-metrics, bridging the threshold-based metrics to field phenological indices results in the highest accuracy (MAE less than 3 days and bias less than 2 days). The strong agreement among the field indices demonstrates the effectiveness of our field phenological surveying approach in generating community-wide forest phenological representation. Our study innovatively scales up the field phenological observations from the individual trees to the species to the community level, and the devised framework enables accurate retrieval of all key phenological events of community-wide, spring canopy development of the forest fragment. [ABSTRACT FROM AUTHOR]

Unmanned Aerial Vehicle (UAV) has seen a dramatic rise in popularity for remote-sensing image acquisition and analysis in recent years. It has brought promising results in low-altitude monitoring tasks that require detailed visual inspections. Semantic segmentation is one of the hot topics in UAV remote sensing image analysis, as its capability to mine contextual semantic information from UAV images is crucial for achieving a fine-grained understanding of scenes. However, in the remote sensing field, recent reviews have not focused on combining "UAV remote sensing" and "semantic segmentation" to summarize the advanced works and future trends. In this study, we focus primarily on describing various recent semantic segmentation methods applied in UAV remote sensing images and summarizing their advantages and limitations. According to the distinction in modeling contextual semantic information, we have categorized and outlined the methods based on graph-based contextual models and deep-learning-based models. Publicly available UAV-based image datasets are also gathered to encourage systematic research on advanced semantic segmentation methods. We provide quantitative results of representative methods on two high-resolution UAV-based image datasets for fair comparisons and discussions in terms of semantic segmentation accuracy and model inference efficiency. Besides, this paper concludes some remaining challenges and future directions in semantic segmentation for UAV remote sensing images and points out that methods based on deep learning will become the future research trend. • Summarize semantic segmentation methods and datasets for UAV remote sensing images. • Provide fair comparisons of representative methods in accuracy and efficiency. • Identify challenges and future directions for UAV image semantic segmentation. • Semantic segmentation methods based on deep learning will be the mainstream. [ABSTRACT FROM AUTHOR]

• For the first time the observations of OCO-2 & 3 was combined to access Xco 2 anomalies. • The Amazon biome has more positive XCO 2 anomalies. • The Brazilian biomes more affected by fire was Amazon, Cerrado and Pantanal. • Xco 2 positive anomalies is correlated with Fire Foci observations. Climate change is a challenge to the global community and one of the causes is the increase of greenhouse gases (GHG) in the atmosphere, especially carbon dioxide (CO 2). The major emission source of this gas into the atmosphere comes from the burning of fossil fuels and biomass burning, on the other hand, the main sink comes from biochemical processes such as photosynthesis. Thus, the observation of CO 2 is a key point to understanding sources and sinks. In this context, The Orbiting Carbon Observatory 2 (OCO-2) and 3 (OCO-3), are a NASA dedicated mission to monitor the column-averaged dry-air mole fraction of carbon dioxide (XCO 2) on a global scale. We combined the OCO-2 and OCO-3 observations to study the spatial distribution of XCO 2 anomalies and how some of these anomalies are related to fire occurrence in the Brazilian Biomes during 2020 and 2021 considering two different seasons, Dry and Wet. The fire occurrence was obtained from Fire Information for Resource Management System (FIRMS) that provides the data from the Moderate-Resolution Imaging Spectroradiometer (MODIS) product of active fires and thermal anomalies at Near-Real Time (MCD14DL, collection 6). The OCO-2/3 observations are affected by cloud formations in wet seasons, we observe that the dry period has more observations. The XCO 2 anomaly values range from ∼ 7.0 ppm to −7.0 ppm and mostly positive anomalies occur in Amazon Biome, and this ecosystem has higher average values for all periods (∼0.9 ppm), compared to the other biomes. The fire occurrence was higher in dry periods, especially in 2020 when unprecedented fire outbreaks were registered in Brazil. The most affected biomes were Pantanal, Cerrado, and Amazon. XCO 2 positive anomalies spatially agree with fire foci over some areas, and the correlation values between them ranged from 0.2 to 0.5 depending on the biome and season, and when considering observations with clouds the correlation is slightly higher. We point out for the first time the possibility of using OCO-2 and 3 combined, also, how positive XCO 2 anomalies are related to fire occurrence in different ecosystems and periods, and the role of cloud detection in this relationship. [ABSTRACT FROM AUTHOR]

The urgency of the biodiversity crisis calls for high-quality habitat data now. Ground-based images (GBIs) represent one of the least-utilized fine-grain data sources with a worldwide coverage and low acquisition cost. The integration of GBIs with other data sources and computer vision can provide objective approaches to improve species–habitat models. We need to understand how GBIs complement existing remote sensing resources, and develop standardized descriptive statistics characterizing the composition and complexity of GBIs. Fine-grained environmental data across large extents are needed to resolve the processes that impact species communities from local to global scales. Ground-based images (GBIs) have the potential to capture habitat complexity at biologically relevant spatial and temporal resolutions. Moving beyond existing applications of GBIs for species identification and monitoring ecological change from repeat photography, we describe promising approaches to habitat mapping, leveraging multimodal data and computer vision. We illustrate empirically how GBIs can be applied to predict distributions of species at fine scales along Street View routes, or to automatically classify and quantify habitat features. Further, we outline future research avenues using GBIs that can bring a leap forward in analyses for ecology and conservation with this underused resource. [ABSTRACT FROM AUTHOR]

Despite significant progress has been made towards crop yield prediction with remote sensing, there exist knowledge gaps on (1) the impacts of temporal resolution of imaging frequencies on yield prediction, (2) transferability of the models among different genotypes and test sites, and (3) translation of these research developments to crop breeding that benefit farmers. Existing research predominantly provides an on-site perspective, frequently missing the complexities of real-world applications. The objectives of this paper are to investigate the transferability and generalization capabilities of yield prediction models for crop breeding across test sites located in North and South Americas. Toward that goal, we tested different machine learning techniques including Random Forest Regressor (RF), Gradient Boosting Regression (GB), and Deep Neural Networks (DNN) for soybean yield prediction with experiments conducted in different climate and growth conditions. A novel transfer learning approach was proposed for genotype selection and categorizing soybean yield for screening high-yield varieties. Furthermore, we studied the effect of temporal resolution on yield prediction, focusing on the critical development stages and optimal aerial survey frequencies for precise yield prediction using large 31,404 sample data. Results demonstrated that the combined dataset of Argentina and United States representing different climate regimes provided the highest performance with an R2 of 0.76 using RF and GB algorithms. The classification approach was proven to be most useful for crop breeding as demonstrated by accurately identifying the high-yielding genotypes. Increasing temporal sampling of key phenological stages significantly improved yield prediction. Although transfer learning yielded promising outcomes across trials within Argentina the efficacy of transferring models from Argentina to the United States was limited, attributed to significant seasonal and climate variations. This study pioneered the use of transfer learning for model adaptability in real-world breeding scenarios, training and transferring models within the South and North Americas, providing actionable insights and strategies for the breeding community, aiming to facilitate improved decision-making for agricultural productivity. [ABSTRACT FROM AUTHOR]

Single object tracking (SOT) in satellite video (SV) enables the continuous acquisition of position and range information of an arbitrary object, showing promising value in remote sensing applications. However, existing trackers and datasets rarely focus on the SOT of oriented objects in SV. To bridge this gap, this article presents a comprehensive review of various tracking paradigms and frameworks covering both the general video and satellite video domains and subsequently proposes the oriented object tracking benchmark (OOTB) to advance the field of visual tracking. OOTB contains 29,890 frames from 110 video sequences, covering common satellite video object categories including car, ship, plane, and train. All frames are manually annotated with oriented bounding boxes, and each sequence is labeled with 12 fine-grained attributes. Additionally, a high-precision evaluation protocol is proposed for comprehensive and fair comparisons of trackers. To validate the existing trackers and explore frameworks suitable for SV tracking, we benchmark 33 state-of-the-art trackers totaling 58 models with different features, backbones, and tracker tags. Finally, extensive experiments and insightful thoughts are also provided to help understand their performance and offer baseline results for future research. OOTB is available at https://github.com/YZCU/OOTB. [ABSTRACT FROM AUTHOR]

Shallow water benthic habitats have been significantly degraded and seriously threatened by intensifying climate changes and anthropogenic stressors. Benthic reflectance (R b λ ) of optically shallow waters (OSWs) is a key parameter for remote sensing of benthic habitats' composition and health status. The remote sensing reflectance (R rs λ ) just above the water surface contains the coupled spectral information of the water column and seabed in OSWs, which makes our effort challenging to separate the signals between seabed and water column properties. Due to too many unknowns of the water column and seabed optical properties, the current approaches are inaccurate or inadequate to retrieve benthic reflectance spectra over a large area from remotely sensed data. In this study, we proposed an enhanced large-scale benthic reflectance (LSBR) retrieval model from satellite data in OSWs with only input of R rs λ . A new spectral library of benthic reflectance was built upon multi-source reflectance datasets obtained from several open-source field measurements. Then, the specific relationship between R b 443 n m and R b 490 n m was developed. By combining a semi-analytical benthic reflectance (SABR) model based on radiative transfer simulations and this relationship, water-column chlorophyll concentration (chl) was determined. Moreover, large-scale water depth was estimated by R rs λ through satellite-derived bathymetry (SDB) technologies. Finally, benthic reflectance was retrieved with the inputs of the satellite-derived R rs λ , chl , and water depth. In this way, the LSBR model fulfills the R b λ retrieval with only input of satellite-derived R rs λ . The new model was successfully applied to Sentinel-2/MSI data in the seagrass region of Xincun Bay and the coral region of Huaguang Reef and further verified by using field measurements and results of the along-track benthic reflectance (ATBR) retrieval model, which combined active lidar and passive high-resolution satellite images. Additionally, to gain insights into the long-term evolution of benthic communities, a typical application of the LSBR model for detecting benthic changes was conducted using eight years of Sentinel-2/MSI data in Heron Reef, depicting a similar consistency between the probable deterioration and restoration region and the sea surface temperature (SST) anomalies. The results demonstrated the robustness and accuracy of the LSBR model in retrieving benthic reflectance at large scales. The LSBR model should be helpful for further investigating benthic changes which can enhance the ability to monitor and assess the health status of submerged ecosystems in OSWs. [ABSTRACT FROM AUTHOR]

Official governmental monitoring of deforestation in the Brazilian Amazon relies on human experts conducting visual analyzes of remote sensing images, an approach that is very expensive and time-consuming due to the enormous geographic extent to be inspected. The main obstacle to the adoption of fully automatic methods is the requirement for the highest possible deforestation detection accuracy, which arises from the official nature of the deforestation reports, used as a basis for deforestation prevention policies and international treaties. This work addresses such a high accuracy constraint by introducing a semi-automatic method that aims at limiting visual analysis to areas where automatic classification is characterized by high uncertainty. We hypothesize that classification uncertainty is correlated with classification error. Therefore, by trusting automatic detections with low uncertainty and subjecting areas of high uncertainty to visual audit, the amount of human intervention in the whole process could be limited to a small fraction of the forest area, while very high accuracies would be achieved. In this work we investigate various uncertainty estimation strategies, including multiple and single inference approaches, and evaluate different uncertainty metrics. Additionally, we conducted a comprehensive analysis to identify the natural factors and deforestation practices that contribute to high uncertainty in automatic classification. We draw interesting conclusions from the analysis of selected regions of two distinct target sites, which exemplify common causes of uncertainty and indicate how uncertainty may be related with ongoing forest degradation processes. Through experimental evaluation at those sites, we demonstrate that the proposed semi-automatic methodology can achieve impressive accuracies in terms of F1-score values, exceeding 97%, while reducing the total human audit workload to just 3% of the target area. • Introduction of a semi-automatic method, significantly reducing human intervention. • An assessment of different uncertainty estimation methods for deforestation mapping. • An empirical investigation of deforestation practices in high uncertainty areas. • An evaluation of the proposed method in two sites in the Brazilian Amazon rainforest. [ABSTRACT FROM AUTHOR]

[Display omitted] Field path planning provides the basis for the autonomous navigation of agricultural vehicles. Existing path planning approaches are constrained to a 2D plane, disregarding the impact of terrain factors on navigation tasks. Furthermore, these methods generate non-global paths, as evidenced by their inability to accommodate the headland turning area while performing path planning based on the actual crop growth in the farmland. Low-altitude remote sensing technology, characterized by its high spatial resolution, timely data acquisition, and strong terrain perception, holds great potential for application in autonomous navigation. This study aims to integrate low-altitude remote sensing technology with path-planning tasks for agricultural vehicles by constructing oblique photography models of fields. The objective is to achieve global 3D path planning leveraging advanced methodologies rooted in deep learning and image processing. Four main steps were included in the proposed method. Low-altitude remote sensing models of field blocks were constructed first. Secondly, the models were converted to image patches for dataset establishment. Thirdly, primary crop row identification was conducted by semantic segmentation. Finally, global 3D paths covering the farmland and headland areas were generated. Tea fields in hilly areas were used to test the algorithm. Experiments revealed that the proposed method could be adapted to different field shapes, row numbers, row spacing, and vehicle turning radii. The proposed method uses deep learning and image processing as the primary technical tools but goes beyond traditional crop row detection to form a global path-planning strategy. In addition, the elevation information enabled by the 3D detection manner allows agricultural vehicles to gain a comprehensive understanding of both their own position and that of their destination. The dataset and the code are available at: https://github.com/ZeroHeading/Global-3D-path-generation.git. [ABSTRACT FROM AUTHOR]

Despite the significant negative impact of marine pollution on the ecosystem and humans, its automated detection and tracking from the broadly available satellite data is still a major challenge. In particular, most research and development efforts focus on one specific pollutant implementing, in most cases, binary classification tasks, e.g., detect Plastics or no Plastics , or target a limited number of classes, such as Oil Spill , Look-alikes and Water. Moreover, most developed algorithms tend to operate successfully only locally, failing to scale and generalize adequately towards operational deployments. Our aim is to address these challenges by introducing a holistic approach towards marine pollutant detection using remote sensing. We argue that constructing such operational solutions requires detectors trained and tested against different types of pollutants, various sea surface features and water-related thematic classes. We offer such a Marine Debris and Oil Spill (MADOS) dataset, composed of high-resolution multispectral Sentinel-2 (S2) data, consisting of 174 scenes captured between 2015 and 2022, with approximately 1.5 M annotated pixels , which are globally distributed and collected under various weather conditions. Moreover, we propose a novel Deep Learning (DL) framework named MariNeXt, based on recent state-of-the-art architectural advancements for semantic segmentation, which outperforms all baselines by at least 12 % in F 1 and mIoU metrics. The extensive quantitative and qualitative validation justifies our choices and demonstrates the high potential of the proposed approach. We further discuss the underlying discrimination challenges among the competing thematic classes. Our dataset, code and trained models are openly available at https://marine-pollution.github.io/. [ABSTRACT FROM AUTHOR]

Hyperspectral target detection is challenging in scenarios where spectral variability is high due to noise, spectral redundancy, and mixing. In addition, this spectral variability also creates the need for target detection algorithms to be robust against variations in the detection threshold. To overcome these challenges, this paper proposes a novel two-stage process for improved target detection in hyperspectral data. In the first stage, coarse detection is performed using a detector with a high probability of detection to identify background samples. These background samples are then used for background learning using an adversarial autoencoder (AAE) network, having spectral angle mapper (SAM) and Huber loss functions to minimize the impact of target pixels' contamination. In the second stage, an inference is made using the spectral difference between the hyperspectral data and the output of the learned background model, which helps in reducing the false alarm rate of the first stage. The proposed approach is compared with seven other target detection techniques using multiple datasets and evaluated through several metrics, such as the area under the curve (AUC) and signal-to-noise probability ratio (SNPR). Results reveal that the proposed technique outperforms other detectors in terms of SNPR, indicating improved target detectability, background suppressibility, and more tolerance to variations in the detection threshold. [ABSTRACT FROM AUTHOR]

• Improved lorenz chaotic systems with higher complexity. • Pixel permutation method based on the image pyramid structure. • Secure and efficient image encryption method. • Galois field multiplication for efficient diffusion of pixels. This paper proposes an improved Lorenz chaotic system and a secure and efficient image encryption method to enhance encryption effectiveness in encrypted images. The proposed improved Lorenz chaotic system addresses the problem of applying the Lorenz chaotic system to image encryption, resulting in weak chaotic characteristics and susceptibility to reconstruction. Dynamic analysis, sensitivity analysis, and randomness testing demonstrate that the improved Lorenz chaotic system exhibits hyperchaotic characteristics, with a maximum Lyapunov exponent of 2.9897. Based on the improved Lorenz chaotic system, this paper proposes an image encryption method that combines image pyramid structure permutation and Galois field diffusion. Unlike most of the current permutation methods limited to a single image layer, this paper proposes a multilayer permutation method based on the image pyramid structure to enhance the permutation effect of image encryption. Although diffusion based on Galois field multiplication operation is efficient and secure, it is less effective in encrypting pixel points with a pixel value of '0′. To address this issue, this paper incorporates DNA computing into diffusion based on Galois field operations, enabling even pure black images to achieve better encryption effectiveness. Experimental results demonstrate that the encryption method proposed in this paper effectively conceals information contained in the plain image. The global Shannon entropy of the encrypted Lena image can reach 7.9975, indicating a high level of randomness and complexity. Notably, even a slight alteration, such as changing a single pixel, results in a significant divergence, with 99.6307 % of the cipher image's pixels being distinct. Moreover, it effectively withstands analysis from various attacks. Therefore, the encryption method proposed in this paper can be effectively applied to grayscale image encryption scenarios requiring relatively high security and encryption efficiency, such as remote sensing image encryption and personal privacy image encryption. [ABSTRACT FROM AUTHOR]

Remote sensing is a cost-effective method for monitoring chlorophyll- a (Chl- a) concentration, an indicator of eutrophication, due to its spatiotemporal effectiveness and availability of historical data. However, its application in shallow, small water bodies poses challenges due to the need for high spatial and temporal resolutions, significant optical interference from other water constituents, and depth-related accuracy issues. This study assessed the performance of Landsat-8 and Sentinel-2 satellites in estimating Chl- a concentrations in Chitgar Lake, Tehran. Semi-empirical models, constructed with 73 and 122 common data pairs between observed data and Landsat-8 and Sentinel-2 reflectance data, respectively, yielded accurate estimates. Two Band (2BAND) algorithm (green and red band ratio, R2 = 0.8, RMSE (root mean square error) = 1.12 µg.L-1 and NRMSE (normalized root mean square error) = 12.4 %) was identified as the best Landsat-8 based model. Both Normalized Difference Chlorophyll Index (NDCI) algorithm (red and red edge bands composition, R2 = 0.82, RMSE = 1.29 µg.L-1, NRMSE = 7.8 %) and 2BAND algorithm (red and red edge band ratio, R2 = 0.81, RMSE = 1.28 µg.L-1, and NRMSE = 7.7 %) were the best models for Sentinel-2 data, with a Power of 0.8. Overall, remote sensing data from both satellites demonstrated appropriate performance and can accurately estimate Chl- a concentrations in Chitgar Lake as a shallow freshwater body. [ABSTRACT FROM AUTHOR]

• LULC changes of three decades from 1991 to 2021 were examined and predicted. • Multi-spectral remote sensing data and RFT were utilized to generate LULC maps. • ACR were considered to analyze the categorical land use change detection. • Impervious land has expanded 10 times within 30 years and will increase in the future. The integration of multi-spectral remote sensing and GIS-based analysis is significant for studying land cover changes, providing valuable insights for informed land management and sustainable development. The present study aims to examine land use land cover (LULC) changes of three decades from 1991 to 2021 and predict the future LULC change in Binh Duong province, Vietnam to explore a future research direction on land use change and associated challenges in the study region. Multi-spectral remote sensing data and random forest tree (RFT) were utilized to generate LULC maps. Areal statistics and annual change rate were considered to analyze the categorical land use change detection. Statistical measures such as user's accuracy, producer's accuracy, kappa coefficient, and confusion matrix were employed to assess the accuracy of LULC classification. To predict future LULC and simulate the spatio-temporal change, we considered previous year LULC maps, independent spatial variables and a combined artificial neural network (ANN) multi-layer perceptron approach. The analysis revealed that there was a huge transition from agricultural lands to residential land with industry and commerce which resulting an expansion of impervious lands and a rapid decline of agricultural land as well as of scrub land and barren lands, and a changeability of forest and plantation, croplands, and waterbodies. Our study revealed that the impervious land has expanded 10 times within 30 years and will increase in the future. [ABSTRACT FROM AUTHOR]

The melting of Antarctic glaciers has become a significant issue as a result of global climate change. Algae on the Antarctic ice/snow is an important part of terrestrial photosynthetic organisms. Monitoring and tracking these algal blooms is crucial for understanding the melting of glaciers in the region. Due to the climatic and natural conditions of the region, traveling to and arranging logistics for monitoring and observing snow algae in the Antarctic continent becomes extremely challenging. To overcome these challenges, a novel algorithm has been developed and designed to automatically detect and analyze green algae (Chlorella sp.) from satellite images. Leveraging the vast and free available data from the Sentinel-2 satellite, the algorithm utilizes its high spectral resolution capabilities, capturing invaluable information from various spectral bands. The algorithm was formulated based on the image obtained on February 28, 2017, where green algae formations were intensively seen in the Ryder Bay. The algorithm was developed based on rule-based detection of algae, with the usage of reflection values from the areas where ground truth was established on this date. The developed algorithm was coded and tested using Python version 3.9. The accuracy analysis of the algorithm was conducted using overall accuracy (OA), F1 score, and Kappa statistical test. As a result of the analysis, the OA, F1 score, and Kappa statistic values were calculated as %91, %88.82-%95.27, and 0.901, respectively. [ABSTRACT FROM AUTHOR]

• A multimodal computational approach was used for the 3D survey of a panel painting. • Data from close-range photogrammetry and laser microprofilometry were fused. • The warping arrow before and after restoration was quantified on the 3D point cloud. • Multi-temporal monitoring of the painting is computed via Cloud-to-Cloud distance. • Micrometric roughness of paint layers is measured by laser microprofilometry. This paper reports the results obtained from the workflow designed for the 3D survey of a Renaissance panel painting using close-range photogrammetry and conoscopic laser microprofilometry, before and after its restoration. The artwork under examination is the central panel of a Renaissance polyptych realized by a late Gothic painter active in central Italy in the late 15th century. This painting has had a rather troubled conservation history following numerous adverse events that have severely damaged it structurally. In preparation for restoration work, the fusion of point cloud data derived from active and passive sensors served to create a single multi-resolution model on which to conduct advanced metrological analyses for multi-temporal monitoring of the changes undergone by the panel painting during and after structural restoration. [ABSTRACT FROM AUTHOR]

• Extract ancient city wall sites at pixel level from LiDAR remote sensing data based on deep learning. • Improvement strategies in the dataset generation, model training, and post-processing steps. • A significant source of inspiration for the inspection, conservation, and monitoring of cultural heritage. Ancient city walls, one of the most notable distinguishing features of Chinese ancient cities, are military defenses constructed of rammed earth. The ancient city walls have considerable study value because they served as the city's boundary and a symbol of power. However, as a result of natural erosion and human activities, many sites have been ruined. Existing optical remote sensing technologies, LiDAR point cloud processing algorithms, and deep learning methods are inadequate for the extraction and segmentation of ancient city wall sites. The novel semantic segmentation method for ancient city wall sites is described in this paper that extracts sites at the pixel level from LiDAR remote sensing data based on deep learning. To begin, the point cloud data collected by airborne laser scanning is processed into DEM data, and the distribution of ancient city walls in the study area is obtained through archaeological survey and expert interpretation. The dataset for deep learning semantic segmentation is then generated using image cropping and data augmentation techniques. Third, implement a U-Net semantic segmentation framework for microtopographic archaeological sites, and predict ancient city wall sites in the testing region after model training. Finally, the deep learning results are optimized using the connected component analysis method, and prediction mistakes such as holes and noise are removed. Taking Jinancheng, the capital city of the Chu kingdom, as an example, the proposed method process can identify and extract the ancient city wall sites at the pixel level, where the evaluation metrics reach 94.12% (Precision) and 81.38% (IoU). The experiment results are excellent due to improvement strategies in the dataset generation, model training, and post-processing steps. Thus, this study is significant for the current survey and protection of ancient city wall sites. The source code will be freely available at https://github.com/wshunli/Open-CHAI-CityWalls. [ABSTRACT FROM AUTHOR]

[Display omitted] The increasing availability of remote sensing imagery led to the emergence of numerous burned area mapping products and methodologies, thus, the necessity of spatial and temporally validating them. The main objective of this paper is to conduct a global analysis and comparison of MODIS and VIIRS active fire dates with dating provided by moderate resolution burned area mapping products. As reference burned area maps, we used the Burned Area Reference Database (BARD) to analyze and compare the performances of MODIS and VIIRS active fires. For 316 path-rows, we calculated the temporal gap between the dates of BARD pixels and those of the five nearest active fires. Olson terrestrial biomes were taken into account in the analysis of results. The agreement rates between the dates of burned areas and of active fires varied based on their geographical distance, fire size and biome. The highest agreement rates were observed for smaller distances (<500 m) and larger fires (>100 ha). Boreal and Mediterranean biomes exhibited the highest values of date percent agreement between burned area pixels and active fires, while the lowest values were found in biomes encompassed by grasslands and savannas. [ABSTRACT FROM AUTHOR]

[Display omitted] Efficient and timely mapping and monitoring of marine vegetation are becoming increasingly important as coastal habits have experienced significant declines in spatial extent due to climate change. Recent advances in machine learning and cloud computing, such as Google Earth Engine, have demonstrated that online analysis platforms make global-scale habitat mapping and monitoring possible. However, the mapping and monitoring of marine ecosystems with remote sensing is challenging, and we lack reliable, generalisable and scalable indices such as NDVI to assess spatiotemporal change in marine habitats. Here, we present a novel method for mapping coastal marine habitats using a kernelised aquatic vegetation index (k NDAVI) with spatially balanced in-water ground truthing and compare it to existing indices and mapping approaches. The kernelised vegetation index provides a simple, consistent, scalable and accurate method for mapping shallow marine vegetation across ∼ 400 km of coastline along mid-west Australia (31.58°S − 29.56°S). This region has significant coastal macroalgae cover that provides critical recruitment habitat for many invertebrates, including commercially valuable fisheries, and has experienced significant loss due to climate-induced heatwaves. We extensively validate k NDAVI and satellite-derived covariates for their utility in mapping SAV using three approaches 1] cross-validation, 2] block cross-validation and 3] site validation. Habitat models that included the kernelised vegetation index achieved excellent agreement (Accuracy > 0.90 and Cohen's kappa > 0.80) for classifying submerged vegetation. We demonstrate that the k NDAVI index has considerable potential for large-scale vegetation monitoring and provides an applicable metric to map spatiotemporal dynamics and more effectively manage these changing coastal habitats. [ABSTRACT FROM AUTHOR]

Accurate large-scale crop yield estimation under climate variability is essential to understanding the dynamics of global food security. The deep learning method has shown well performance for crop yield estimation because of its high capacity for temporal pattern recognition. However, most existing deep learning models were usually based on multi-source time series with weekly or coarse temporal resolutions, which simplified the dynamics of crop growth progress. How to effectively learn the long-term and short-term crop response from the remotely sensed and agro-meteorological time series at daily intervals remains challenging. In this study, a hierarchical deep learning model, named Daily DeepCropNet (DDCN), has been developed to process daily time series of satellite-derived vegetation indices and climatic variables for corn yield estimation. The DDCN model was a three-level pyramid structure that incorporated the corn phenology durations to extract crop growth patterns from the daily inputs. The Transformer and Long Short-Term Memory (LSTM) were implemented as a temporal filter in each level to capture the short-term crop growth patterns. The temporal patterns were hierarchically transferred from weekly to growing season aggregation to learn long-term dependency. The experiment was conducted at county level over the US Corn Belt from 2006 to 2020. LSTM, Transformer, and Random Forest (RF) models were built for comparison. The input analysis showed that the daily time series of climatic variables provided more temporal fluctuations than weekly aggregated inputs. The DDCN model provided the highest accuracy for end-of-season yield estimation under normal climate conditions in 2018–2020 (RMSE = 0.86 Mg ha−1) and extreme heat and drought conditions in 2012 (RMSE = 1.17 Mg ha−1). Further analyses indicated that the DDCN model had the potential for capturing the short-duration stresses. We also found that the DDCN model identified the silking phase as the key growth stage and achieved higher accuracy than other models for in-season yield estimation. The results showed that features extracted by the DDCN model had higher correlation coefficients with corn yield for the upper levels. By evaluating the contributions of different variables, we found that vegetation indices yield higher impact than climatic variables for yield estimation under both normal and stressful climate conditions. This study provided a promising approach to extracting the temporal pattern of crop growth at daily time steps. The DDCN model could be further integrated with the process-based model to achieve the fusion of knowledge and data for explainable AI in agricultural system modeling. [ABSTRACT FROM AUTHOR]

Remote sensing image scene classification (RSI-SC) is crucial for various high-level applications, including RSI retrieval, image captioning, and object detection. Deep learning-based methods can accurately predict scene categories. However, these approaches often require numerous labeled samples for training, limiting their practicality in real-world RS applications with scarce label resources. In contrast, few-shot remote sensing image scene classification (FS-RSI-SC) has garnered substantial research interest owing to its potential to mitigate the need for extensive training samples. In recent years, there has been a surge in studies on FS-RSI-SC. This paper presents a comprehensive overview of FS-RSI-SC research, categorizing existing methods into two groups. The first group comprises approaches based on data augmentation, transfer learning, metric learning, and meta-learning. Our analysis reveals that most existing FS-RSI-SC methods fall into the meta-learning category, employing attention mechanisms, self-supervised learning (SSL), and feature fusion techniques for enhanced performance. Additionally, transfer learning-based methods consistently outperform other approaches in this category. The second group is centered around large-scale pre-training, which has demonstrated remarkable competitiveness across various tasks, including FS-RSI-SC. This special group of methods has shown considerable potential and is expected to attract more attention with the increasing popularity of large-scale pre-training and the unimodal and multimodal foundation models. Moreover, we proposed a pipeline that harnesses the capabilities of powerful large vision-language models (VLMs) as image encoders, establishing new baselines for FS-RSI-SC on commonly used datasets under standard experimental settings. Our empirical results validated the effectiveness of utilizing large VLMs and highlighted their potential for FS-RSI-SC. Through a joint analysis of state-of-the-art methods and our experiments with VLMs, we identified the prevailing challenges in FS-RSI-SC and outlined promising directions for future research. [ABSTRACT FROM AUTHOR]

Geographic spaces is a vast and complex system involving multiple elements and nonlinear interactions of these elements, and rich in geographical phenomena, processes and patterns. Artificial intelligence methods (AI) are increasingly utilized to extract information of interest, patterns and insights from massive remote sensing (RS) images. Among them, two representative paradigms for RS interpretation are knowledge-driven symbolism and data-driven connectionism. Knowledge-driven approaches are certain and theoretically sound, which consider the regularities, formulae with accompanying theorems, and expert knowledge. However, it is difficult to address huge-scale RS data and complex nonlinear problems. The data-driven paradigm has stronger data learning and representation ability, yet poor interpretability, low trustworthiness of interpretation results, and reliance on massive labeled data. To address these limitations, we argue that current RS intelligent interpretation requires the guidance of geographic ideas and a unified theoretical framework rather than the independent use of remotely sensed data, duplicative development of interpretation models, or isolated applications of geographic information extraction. Thus, we introduce the concepts of trustworthy RS interpretation (TRSI) with multi-granularity of space, time, and attribute to facilitate the understanding of the morphology, types, and indicators (visual perception) and the perspective of the structure, evolution, and trends (geographical cognition) of complex geographical space. Second, integrated technologies of the TRSI with multiple scales and stages are developed, including a pixel-level visual perception, a geo-parcel-level quantitative inversion, and a scene-level geographical cognition. Finally, a RS big data interpretation system with quantitative semantic parsing is designed to support precise applications with cloud–edge-end collaboration. Rather than a simple overview and accumulation of traditional RS intelligent interpretation methods, this paper analyzes current challenges in RS interpretation and proposes new concepts, technologies, and applications of TRSI. It will empower us to perceive the complex geographical space comprehensibly and reveal its structural composition and evolutionary mechanism, serving geospatial decision applications such as evaluation, planning, and forecasting. [ABSTRACT FROM AUTHOR]

Automated extraction of polygonal building contours from high-resolution remote sensing images is important for various applications. However, it remains a difficult task to achieve automated extraction of polygonal buildings at the level of human delineation due to diverse building structures and imperfect image conditions. In this paper, we propose Line2Poly, an end-to-end approach that uses feature lines as geometric primitives to achieve polygonal building extraction by recovering topological relationships among these lines within an individual building. To extract building feature lines with precision, we adopt a two-stage strategy that combines Convolutional Neural Network (CNN) and transformer architectures. A CNN-based module extracts preliminary feature lines, which serve as positional priors for initializing positional queries in the subsequent transformer-based module. For polygonal building contour reconstruction, we devise a learnable polygon topology reconstruction module that predicts adjacency relationships among discrete lines, and integrates lines into building polygons. The resultant building polygons, based on feature lines, exhibit inherent regularity that aligns with manual labeling standards. Extensive experiments on the Vectorizing World Buildings dataset, the WHU aerial building dataset and the WHU-Mix (vector) dataset validate Line2Poly's impressive performance in building feature line extraction and instance-level building detection. Moreover, Line2Poly's predictions exhibit the highest level of concurrence with manual delineations, with over 83% agreement on the WHU aerial building test set and 68.7/59.7% on the WHU-Mix (vector) test set I and II, respectively. [ABSTRACT FROM AUTHOR]

The extraction of building footprints, as a highly challenging task in remote sensing (RS) image-based geospatial object detection and recognition, holds significant importance. Due to the strong coupling in RS images between the body and boundary of buildings, the ability of most currently advanced deep learning models in building footprint extraction remains limited, inevitably meeting the extraction performance bottleneck. To this end, we propose a novel High-resolution Decoupled Network, HD-Net for short, for precious building footprint extraction in RS. HD-Net follows the well-known high-resolution network (HRNet) architecture, which can to a great extent alleviate the coupling issues between body and boundary using its multi-scale information interaction in parallel. More specifically, Our HD-Net innovatively designs the multiple stacked multi-scale feature fusion (MFF) modules, where the MFF module is performed by combining the deep supervision technique and a feature decoupling–recoupling (FDR) module. The FDR module adeptly untangles coupled features into two distinct elements: body and boundary, yielding feature maps enriched with semantic information. This configuration facilitates a step-wise refinement of building extraction and boundary predictions, ensuring the overall continuity of buildings and the precision of their boundaries. Experiments conducted on the three widely-used building datasets, i.e., Massachusetts, WHU, and Inria, demonstrate that HD-Net achieves the most competitive results with minimal parameter count. In detail, HD-Net outperforms contour-guided and local structure-aware network (CGSANet) with intersection over union (IoU) improvements of 0.40%, 0.95%, and 0.73% on the mentioned datasets, while CGSANet is the state-of-the-art algorithm using a hybrid loss function and deep supervision strategy. Furthermore, the code of the HD-Net will be made available freely at https://github.com/danfenghong/ISPRS_HD-Net. [ABSTRACT FROM AUTHOR]

The state-of-the-art methods for cloud detection are dominated by deep convolutional neural networks (DCNNs). However, it is very expensive to train DCNNs for cloud detection and the trained DCNNs do not always perform well as expected. This paper proposes a shallow CNN (SCNN) by removing pooling/unpooling layers and normalization layers in DCNNs, retaining only three convolutional layers, and equipping them with 3 filters of 1 × 1 , 1 × 1 , 3 × 3 in spatial dimensions. It demonstrates that the three convolutional layers are sufficient for cloud detection. Since the label output by the SCNN for a pixel depends on a 3 × 3 patch around this pixel, the SCNN can be trained using some thousands 3 × 3 patches together with ground truth of their center pixels. It is very cheap to train a SCNN using some thousands 3 × 3 patches and to provide ground truth of their center pixels. Despite of its low cost, SCNN training is stabler than DCNN training, and the trained SCNN outperforms the representative state-of-the-art DCNNs for cloud detection. The same resolution of original image, feature maps and final label map assures that details are not lost as by pooling/unpooling in DCNNs. The border artifacts suffering from deep convolutional and pooling/unpooling layers are minimized by 3 convolutional layers with 1 × 1 , 1 × 1 , 3 × 3 filters. Incoherent patches suffering from patch-by-patch segmentation and batch normalization are eliminated by SCNN without normalization layers. Extensive experiments based on the L7 Irish, L8 Biome and GF1 WHU datasets are carried out to evaluate the proposed method and compare with state-of-the-art methods. The proposed SCNN promises to deal with images from any other sensors. The code and data will be open for further research. 1 1 https://github.com/dfchai/SCNN. [ABSTRACT FROM AUTHOR]

Crop mapping provides information on crop types and cropland spatial distribution. Therefore, accurate and timely crop mapping serves as the fundamental step to higher-level agricultural applications, such as crop yield prediction. Recently, deep learning (DL) classification models have been explored for crop mapping. However, most existing methods are still limited in practical applications for large-scale crop mapping due to the great need for representative labeled data that cover diverse geographical conditions. To address this issue, we proposed a novel deep active learning (AL) method, named Bayesian Semi-Supervised Active Learning (BSSAL), using time-series Sentinel-2 imagery for large-scale crop mapping. This framework consists of three parts: 1) a Bayesian neural network (BNN) for crop classification, 2) a Semi-Supervised task to leverage massive satellite imagery without crop type information, and 3) an AL strategy to select the most informative samples to be labeled. The proposed framework was validated on five crop classes over six sites with different geographical conditions across the U.S. in testing years 2019–2021. The BSSAL framework produced a mean OA of 98.69% in the end-of-season experiment and significantly outperformed other comparison methods. Moreover, it achieved an average OA of 97.00% with in-season satellite imagery from the day of the year (DOY) 140 (early April) to 200 (mid-July) and greatly improved the timeliness of crop mapping. Furthermore, the BSSAL framework had the best performance in all study sites, showing the robustness of the global model on local testing. By monitoring the accuracy of each crop type along with the number of samples selected, the results showed the effectiveness of the proposed active data query strategy to random choice. Overall, this study provided a robust framework for large-scale crop mapping with less labeling budget and higher classification accuracy. The source code is available at: https://github.com/YXu556/BSSAL. [ABSTRACT FROM AUTHOR]

Understanding the spatial–temporal dynamics of both natural and planted forests is critical for sustainable forest management and assessing ecological benefits or impacts. In 2020, China's forest area encompassed approximately 220 million hectares, accounting for around 5 % of the global forest area. Furthermore, China boasts the largest planted forest area worldwide. However, knowledge regarding the spatial–temporal dynamics of China's natural and planted forests has remained elusive. To fill this gap, we first generated wall-to-wall natural and planted forest maps at 30-m resolution for China every-five years from 1990 to 2020 by using 447,730 images from Landsat-4/5/7/8/9 surface reflectance archive and 656,920 field samples, which were a combination of crowdsourced data and expert knowledge. Secondly, we analyzed the spatial–temporal dynamics of natural and planted forests at multiple-scales from 1990 to 2020. The resultant maps achieved an overall accuracy ranging from 77.33 %±0.67 % to 81.78 % ± 0.59 % and revealed opposite trends in the areas of natural and planted forests from 1990 to 2020. During these three decades, China's planted forest area increased by 447,500 km2 while the natural forest area decreased by 219,100 km2. The spatial–temporal dynamics of China's natural and planted forests result from China's vast forest management programs in addition to social and economic factors. Variations of these factors at different scales not only determined the primary goals of protecting natural forests and expand planted forests but also resulted in increasing forest areas for many consecutive years by planting forests that offset the loss of natural forests. Our maps provide timely and valuable insights into the benefits and impacts of natural and planted forests in China, which are both closely linked to sustaining human well-being. [ABSTRACT FROM AUTHOR]

Land-use/land-cover (LULC) classification of high spatial resolution (HSR) remote sensing imagery has been successfully improved using deep learning techniques. However, the current deep learning-based classification methods necessitate the division of remote sensing imagery into smaller and fixed image patches, primarily due to computational constraints arising from the extensive size of these images. This approach limits the receptive field of the classification network and hinders the handling of different-scale LULC objects. A key problem is how to automatically select the appropriate scale of patch for different objects with a deep learning network. To address this challenge, a scale-aware classification network (SAN) based on deep reinforcement learning (DRL) is proposed. In SAN, the state of each image patch is represented by a reduced-resolution version of the high-spatial-resolution (HSR) remote sensing image, referred to as a 'thumbnail', and a positional encoding. The scale selection actions are performed by a scale control agent. A feature indexing module is also proposed to enhance the ability of the agent to distinguish the location of the current image patch. The action switches the patch scale and the viewing area of context branch of a two-branch classification network, which extracts and fuses the features of the multi-scale images. The SAN framework adjusts the network parameters to perform the appropriate scale selection action based on the mapping reward received for the selected scale. In this way, the SAN framework is able to introduce more appropriate contexts by adjusting the scale of the network input based on RL, without the need for labeled scale selection samples. The experimental results obtained using two publicly available datasets and a newly built dataset demonstrate that SAN outperforms the previous LULC deep learning methods with fixed patches, particularly for large-scale mapping applications. When compared to state-of-the-art approaches such as GLNet and WiCoNet, which combine global and local information for segmentation, as well as CascadePSP and MagNet, renowned for their progressive segmentation capabilities, SAN consistently demonstrates approximately 10% higher accuracy. The codes for this research are openly available at http://rsidea.whu.edu.cn/resource_sharing.htm. [ABSTRACT FROM AUTHOR]

Land-cover information reflects basic Earth's surface environments and is critical to human settlements. As a well-established deep learning architecture, the fully convolutional network has achieved impressive progress in various land-cover mapping tasks. However, most research has focused on designing powerful encoders, ignoring the exploration of decoders. The existing handcrafted decoders are relatively simple and lack flexibility, limiting the generalizability for complex remote sensing scenes. In this paper, we propose a Land-cOVEr mapping Neural Architecture Search framework (LoveNAS) to automatically find efficient decoders that are compatible with the encoders and tasks. Specifically, LoveNAS introduces a hierarchical dense search space, including densely connected layer-level and multi-scale operation-level search spaces. The search spaces contain independent connection and operation fusion strategies, facilitating sufficient interaction of multi-scale features. After searching based on large-scale datasets, a series of pre-trained encoders and adaptive decoders are obtained. These can be smoothly applied to multi-scene tasks using weight-transfer network training. Experimental results on normal and disaster scenes shows that LoveNAS outperforms 16 handcrafted architectures and NAS methods. Some searched structures coincide with the existing advanced artificial designs, revealing the potential value of LoveNAS in network design and guidance. Group's website: http://rsidea.whu.edu.cn/resource%5fsharing.htm. GitHub page: https://github.com/Junjue-Wang/LoveNAS. [ABSTRACT FROM AUTHOR]

• Fe/Mg phyllosilicates and mono- and polyhydrated Mg sulfates are detected in Mikumi crater. • Possible desiccation/hydraulic fracturing on crater floor. • Phyllosilicates are on top of the sulfate units, meaning that they formed later than sulfates. • Mikumi crater is a key area in Meridiani Planum to investigate Mars' ancient climate. The region of Meridiani Planum on Mars retains the evidence of multiple surface and near-surface aqueous processes that span from the Noachian to the Hesperian. This makes it an interesting spot for better understanding the characteristics and the evolution of Mars' aqueous environment between these two fundamental epochs. Here we investigate the mineralogy, stratigraphy, and morphology of the Noachian–Hesperian sedimentary materials that fill a 16-km-wide crater named Mikumi, located in the northern part of Meridiani Planum. The analysis was carried out combining observations from the Compact Reconnaissance Imaging Spectrometer for Mars (CRISM), the High Resolution Imaging Science Experiment (HiRISE), and the Context Imager (CTX). About 20–30 meters, on average, of Fe/Mg phyllosilicate-rich rocks and a thick sequence (75 m) of both mono- and polyhydrated Mg sulfates are detected on the crater floor. Both layers postdate the impact event. Most of the hydrated terrains are extensively polygonally fractured, suggesting the possible action of water-related processes such as desiccation and hydro fracturing on the sediments. Stratigraphic analysis clearly shows that Fe/Mg phyllosilicates were deposited on top of the sulfate sequence, meaning that they are younger than sulfates. This can have interesting implications on paleoclimatic studies of Mars as it shows that a simple timeline going from a clay-rich Noachian to a sulfate-rich Hesperian oversimplifies the evolution of Mars' mineralogy and related aqueous environment at the Noachian–Hesperian boundary. [ABSTRACT FROM AUTHOR]

• Algorithm to model distributed passive microwave interferometers. • Study of the reduction of side-lobes levels with hexagonal antenna shape. • Modelling of distributed mission in low Earth Orbit and control methodologies. • Assessment of the effect of the control error of the spacecraft's relative position on the interferometer performances. • Application of the methodology to an innovative mission concept for high-resolution passive L-band observations. The impact of formation flying on interferometry is growing over the years for the potential performance it could offer. However, it is still an open field, and many studies are still required. This article presents the basic principles behind interferometry focusing first on a single array and secondly on a formation of satellites. A sensitivity analysis is carried out to evaluate how the performance of the interferometry is affected by an error in the relative position in the formation geometry. This is estimated by computing the loss of the performance in terms of percentage deviation due to a non-nominal relative trajectory, including two-dimensional errors and defining a payload index. The main goal of this study is to estimate whether some errors in the relative state are more impacting than others. The final objective is to compute the link between a position error and a specific loss of performance, to foresee the origin of the the error. Furthermore, a dynamical model is developed to describe the relative motion in the Low Earth Orbit environment, considering both the unperturbed and the J 2 and drag contributions. A Proportional, Integral and Derivative controller is implemented for the position control of a multiple satellite formation flying, considering a low thrust control profile. The Formation Flying L-band Aperture Synthesis study is taken as the case scenario, analysing both nominal and non-nominal configurations. This study serves as a starting point for the development of a combined tool to assess the performance of the interferometry and the control on the relative state for future remote sensing studies involving relative motion. [ABSTRACT FROM AUTHOR]

New Delhi experienced worst air-quality days and occurrence of severe smog on 03–04 November 2022. Remote sensing data from various platforms, NOAA HYSPLIT and ERA-5 is utilized to monitor the episode of high smog conditions over Delhi during October- November 2022. It has been observed that AOD has increased up to 4 times till 04 November 2022 when compared to before the starting of burning period. Since the anthropogenic source regions are concentrated around Delhi; the adjoining areas have shown increased burning using fire count data from MODIS Aqua. Study revealed that wind pattern with reduced speeds at lower altitudes is favored to sustain smog for an extended period. The study revealed that the smog over Delhi region for an extended period is formed majorly by the transport of particulate matter and CO emission from nearby areas, compounded by the prevailing fog. Air-mass trajectory analysis confirms that atmosphere over Delhi region was influenced by the transported mass of aerosol particles which corroborated well with the areas characterized by the fire counts, indicating a role of residual burning. [ABSTRACT FROM AUTHOR]

Imaging spectrometry is a powerful tool among remote sensing techniques to analyze the chemical compositions and physical properties of the lunar regolith. Recalibration of the classical method byLucey and co-workers with Moon Mineralogy Mapper (M3) data provides an opportunity to understand the FeO, TiO 2, and optical maturity distributions over the lunar nearside on a global scale. We estimated the chemical compositionsand optical maturity from the M3 spectral reflectance around the 1 µm region. The standard chemical compositions of the lunar landing sites were estimated based on origin optimization and data fitting. We obtained an optimized origin of (0.08, 1.18) for FeO and (0.07, 0.71) for TiO 2. The Spectrophotometric evaluations of Catharina and Cyrillus craters, located in the lunar nearside are categorized under the matured category. The estimation results reveal that the FeO wt % of the Catharina crater varies from 0.0002 to 19.62, and that of the Cyrillus crater varies from 0.0001 to 12.32. TiO 2 wt% of Catharina crater varies from 0.01 to 86.12, while Cyrillus Crater varies from 0.01 to 4.34. The optical maturity (OMAT) of the Catharina crater varies from 0.0001 to 0.19 to that of the Cyrillus Crater Varies from 0.01 to 0.17. The overall summary of the results points out that lunar regolith is a fabulous repository of FeO and TiO 2 and quantitatively this metal abundance may be more than the anticipated values. [ABSTRACT FROM AUTHOR]

This review article provides an overview of various aspects of lunar exploration, including missions to the Moon, collection and analysis of lunar sample data in laboratories, and the processing and analysis of remote sensing data, particularly using radar techniques. Both remote sensing and in-situ methods are critical for advancing our understanding of the lunar surface and its properties. This review article focuses on the identification and quantification of water–ice deposits located in areas such as Permanently shadowed areas (PSRs) and the lunar poles (section " Lunar Poles and PSRs: A Special Environment "). These volatile resources have the potential to serve as valuable sources of fuel for future missions, making it crucial for the lunar community to determine their abundance and distribution. After thoroughly examining lunar samples using high-precision laboratory techniques, many preconceptions were dispelled which is highlighted in the section " Laboratory Investigation of Lunar samples ". But as in-situ observations are difficult to acquire, especially terrestrial bodies samples, remote sensing techniques allow the global understanding of the surface. The article specifically highlights the importance of understanding the electrical characteristics of the lunar surface and how radar inversion can provide valuable information in this regard. The " Conclusion " section of this review article serves as a key takeaway for readers, underscoring the critical role that both in-situ and remote sensing techniques play in advancing our understanding of the Moon. [ABSTRACT FROM AUTHOR]

Unmanned aerial vehicle-based hyperspectral imaging (HSI) and machine learning (ML) techniques are applied for crop field data collection such as yield, nitrogen content, leaf chlorophyll, biomass estimation, leaf area index (LAI), and biotic and abiotic stress. It is very difficult to identify with certainty best practices for specific applications and for targeted crop species because the results are frequently contradictory; models or indices sometimes perform well, but other times they do not. ML and HSI are useful and sophisticated techniques, but treating them as first-option methods may be detrimental in the long term. Favoring HSI in situations where a multispectral sensor performs equally well only delays end-user adoption of technology owing to cost and complexity. The past few years have seen increased interest in unmanned aerial vehicle (UAV)-based hyperspectral imaging (HSI) and machine learning (ML) in agricultural research, concomitant with an increase in published research on these topics. We provide an updated review, written for agriculturalists, highlighting the benefits in the retrieval of biophysical parameters of crops via UAVs relative to less sophisticated options. We reviewed >70 recent papers and found few consistent results between similar studies. Owing to their high complexity and cost, especially when applied to crops of low value, the benefits of most of the research reviewed are difficult to explain. Future effort will be necessary to distill research findings into lower-cost options for end-users. [ABSTRACT FROM AUTHOR]

Neural Radiance Fields (NeRF) is a technique for reconstructing real-world scenes from multiple views. However, existing methods mostly focus on the visual effects of scene reconstruction while neglecting geometric accuracy, which is crucial in photogrammetry and remote sensing. In this paper, we propose a method called LiDeNeRF which uses LiDAR point cloud to provide depth priors for NeRF reconstruction. The goal of LiDeNeRF is to achieve real-time rendering of NeRF scenes and 3D reconstruction results with high geometric accuracy. In this method, first, the LiDAR point cloud is projected onto images to generate a sparse depth map. Then, by triangulating the sparse depth and using a multi-view image depth propagation method, a dense depth map with high accuracy is quickly generated as the depth prior for NeRF. Finally, a new depth correction module is designed and embedded into the NeRF rendering pipeline to improve the accuracy of scene depth estimation. The experimental results evince that our methodology has attained paramount performance in both novel view synthesis and 3D reconstruction tasks. Our source code is made available at https://github.com/WPC-WHU/LiDeNeRF. [ABSTRACT FROM AUTHOR]

Domain adaptation methods are able to transfer knowledge across different domains, tackling multi-sensor, multi-temporal or cross-regional remote sensing scenarios as they do not rely on labels or annotations in the target domain. However, most of the previous studies have focused on closed-set domain adaptation, based on the assumption that the source and target domains share identical class labels. Real-world scenarios are typically more complex, and the model could potentially encounter novel classes that are not previously included in the source domain, commonly referred to as "unknown" classes. Here we investigate the open-set domain adaptation scenario in the field of remote sensing scene classification, where there is a partial overlap between the label space of the target domain and that of the source domain. To deal with this problem, we propose a novel open-set domain adaptation method for scene classification using remote sensing images, which is named Multi-Adversarial Open-Set Domain Adaptation Network (MAOSDAN). Our proposed MAOSDAN consists of three main components. First, we employ an attention-aware Open Set BackPropagation (OSBP) to better distinguish the "unknown" and "known" samples for the target domain. Then, an auxiliary adversarial learning is designed for mitigating the negative transfer effect that arises from forcefully aligning the "unknown" target sample in network training. Finally, we adopt an adaptive entropy suppression to increase the probability of samples and prevent some samples from being mistakenly classified. Our proposed MAOSDAN achieves an average score of 75.07% in three publicly available remote sensing datasets, which significantly outperforms other open-set domain adaptation algorithms by attaining 4. 52 ∼ 17. 15 %. In addition, MAOSDAN surpasses the baseline deep learning model with 18.12% improvement. A comprehensive experimental evaluation demonstrates that our MAOSDAN shows promising prospects in addressing practical and general domain adaptation scenarios, especially in scenarios where the label set of the source domain is a subset of the target domain. [ABSTRACT FROM AUTHOR]

Newly constructed areas (NCA) are continuously emerging with the development of urbanization, which, however, triggered a series of ecological and environmental issues. Therefore, monitoring NCA is of great significance for sustainable urbanization. Extraction of NCA from remote sensing images involves semantic segmentation of constructed areas and change detection. However, the acquisition of samples for training change detection models is time-consuming. In this study, we proposed a multi-task learning framework including unsupervised change detection and supervised semantic segmentation for NCA extraction. The main contributions of this study are: (1) A deep multivariate alternation detection (DMAD) algorithm was proposed for unsupervised change detection of hyperspectral images. By introducing the optimization objective of Multivariate Alternation Detection (MAD) in the loss functions, the features of changed areas can be extracted more effectively; (2) A multi-task learning framework was proposed for the joint training of DMAD change detection and supervised semantic segmentation, where the two modules were mutually optimized via semantic masks and consistency loss. Based on Orbita Hyperspectral Satellites (OHS) images, we conducted experiments in two cities (Qinzhou and Wuzhou) of Guangxi province. The experimental results indicated that the proposed method can achieve the F1-score better than existing unsupervised change detection methods (e.g., DCCA and DSFA) by an average of 15%, and the OA better than existing semantic segmentation methods (e.g., U-Net and FreeNet) by an average of 1%. The F1-score and Kappa for NCA extraction reached above 0.80 for both study areas. The implementation of this paper will be made available at https://github.com/tulilin/Multitask_NCA. [ABSTRACT FROM AUTHOR]

Remote sensing imagery usually suffers from information loss issues due to the sensor defects and harsh atmospheric conditions, resulting in a significant decrease in data quality. In light of this, a novel image reconstruction method is proposed to improve the usability of such data. The proposed method uses the isophote information in reference image as constraint to recover the missing information in target image. Isophote refers to the neighboring pixels with same illumination intensity, that is, the pixels with similar gray values. The adaptive weighted gradient is put forward to calculate the isophote of a pixel. Due to the greater weight in the direction of minimum gradient, the difference between similar pixels in image tends to remain constant to minimize the isophote difference between images. As a result, in the reconstructed image, radiometric differences between reference image and target image are attributed to areas where pixel values change significantly, and the pixels belonging to same ground feature will have similar values, thereby ensuring the internal stability of the feature and increasing the variance between different features. Furthermore, to control the color fidelity of ground features and improve the reconstruction accuracy, the target image is updated with certain pixels before isophote-based image reconstruction. These pixels are filled with the auxiliary information outside the missing area under color-structure control, and evenly distributed in each uniformly colored small zone of the missing area. Experiments under different missing information cases have demonstrated the effectiveness of the proposed method, and the experimental results indicate that the proposed method can obtain visually reasonable images with high color consistency and reconstruction accuracy even for images with large missing areas. The source code of this work is available at https://github.com/YuXiaoyu221/IC_CSC-Information-Reconstruction. [ABSTRACT FROM AUTHOR]

Monitoring forest aboveground biomass (AGB) is essential for quantifying the carbon cycle and mitigating climate change. Tropical secondary forests are significant carbon sinks that sequester large amounts of carbon dioxide. While recent studies have attempted to estimate the AGB recovery rates in tropical forests, considerable uncertainty remains in the estimation of AGB recovery of secondary forests and the spatial variability of the effects that different environmental conditions and degrees of human use may have on AGB recovery. These knowledge gaps hinder further understanding of climate change mitigation potential of secondary forests. Remote sensing products provide spatially and temporally explicit information for understanding and monitoring secondary forest dynamics. To explore the local effects of different factors on AGB of secondary forests in Brazil, we used geographically weighted regression (GWR) models that account for spatial heterogeneity in geospatial data to estimate the AGB of secondary forests in Brazil. Secondary forest areas (29142 polygons) were extracted from Brazil's forest age maps between 1984 and 2019. The AGB of these areas was derived from the Climate Change Initiative Biomass maps. The effects of selected predictors such as forest age, climatic water deficit, the cation exchange capacity of soil and surrounding tree cover were analyzed. The two most influential factors, forest age and surrounding tree cover were utilized to estimate the AGB and the recovery rates per year. Our results show the high spatial variation of different predictors' effects on the AGB of secondary forests. Also, the GWR model (with an adjusted R2 of 0.74) showed considerable improvements regarding "goodness of fit" of models compared with the Ordinary Least Squares (with an adjusted R2 of 0.53). Our estimated average AGB recovery rate across all Brazil's biomes is 7.5 Mg ha−1 yr−1 (using forest age) for the first 20 years. We presented the map of the spatial variation of AGB recovery rates in Brazil. The estimated AGB recovery rates range using forest age is 28.9 Mg ha−1 yr−1. Our estimated mean AGB recovery rates of different biomes are 17.7 % on average higher than IPCC default rates. Our results provide baseline information for reducing uncertainties related to carbon sink estimation of secondary forests in Brazil, hence assisting in developing sustainable forest management and ecosystem restoration strategies. [ABSTRACT FROM AUTHOR]

Recently, ensemble multiple deep learning (DL) classifiers has been reported to be an effective method for improving remote sensing classification accuracy. Although these approaches still follow the conventional pattern of inputting instance features and outputting corresponding classes, they often overlook the intrinsic relationships between pixels beyond their spatial features. As a result, the diversity in the ensemble classification results primarily relies on different DL models. However, training the DL models consumes a significant amount of time, and training multiple networks not only incurs additional time costs but also affects the overall efficiency. To address this, a new approach has been proposed in this paper, which takes advantage of the relationships between pixels and their combinations to generate diverse classification results. It's a novel ensemble classification framework, termed as the Doublet-Based Ensemble Classification Framework (DBECF), which eliminates the need for multiple classifiers. The DBECF starts by utilizing the training set to combine different samples to generate doublets. Then, features are assigned to these doublets through an exponentiation operation, resulting in a doublet training set. Using both the original training set and the derived doublet datasets, the DBECF is trained. For each input pixel, the DBECF produces multiple classification results, which are then integrated to obtain a more accurate output. To validate the proposed approach, experiments were conducted on three datasets, including multispectral images, hyperspectral images, and time series images. The maximum accuracies achieved by DBECF on the three datasets are 87.80 %, 97.71 %, and 83.51 %, respectively. In comparison to the contrastive methods, the incremental improvements in accuracy are 3.73 %, 7.66 %, and 9.16 %, respectively. The experimental results indicate that no matter using DL or non-deep learning for training, our proposed framework achieves progress on accuracy improvement outperforming classifications using comparative approach that based on single instance. This research provides a new perspective on the combination of DL and ensemble learning, highlighting its important implications and practical value in enhancing classification accuracy and efficiency. [ABSTRACT FROM AUTHOR]