Search

Precise and timely mapping of winter wheat is essential for food security. Current methods are limited by insufficient training data and a lack of long-term early mapping verification. This research proposes a framework that uses a cascade index to generate high-quality training samples for winter wheat mapping automatically. By considering the phenological characteristics of winter wheat and similar crops, the cascade index method screens and acquires these samples. Combined with a random forest model, mapping was conducted in Henan Province and the Agricultural Statistics District (ASD) 2020 area in the U.S. In Henan, early mapping from 2018 to 2022 assessed differences between model transfer and current-year samples. Results showed that using October-April imagery based on model migration achieved an overall accuracy (OA) of over 90%, while October-February imagery based on current-year samples also exceeded 90%. In some years, early mapping using only October-December data also achieved over 90% OA. These findings demonstrate the proposed model's viability for large-scale early winter wheat mapping using satellite imagery. Furthermore, this method demonstrates adaptability, mapping results achieving over 93.69% OA when transferred to the United States.

An increasing availability of remote sensing data in the era of geo big-data makes producing well-represented, reliable training data to be more challenging and requires an excessive amount of human labor. In addition, the rapid increase in graphics processing unit processing power has enabled the development of advanced deep learning algorithms, which achieve impressive results in the field of satellite image processing. However, they require a huge and comprehensive training dataset to avoid overfitting problems and to represent a generalizable model. Thus, moving toward the development of nonsupervised deep learning (NSDL) models in different remote sensing applications is an inevitable need. To provide an initial response to that need, this article performs a comprehensive review and systematic meta-analysis of recently published research articles focusing on the applications of NSDL for remote sensing data processing. In order to identify future research directions and formulate recommendations, we extract trends and highlight interesting approaches from this large body of literature. Consequently, current challenges, prospects, and recommendations are also discussed to uncover the trend. According to the results, there is a sharp increasing trend in the applicability of NSDL methods during these few years particularly, with the advent of new deep architectures, such as adversarial, graph, and transformer models. As a result, this review article discusses different remote sensing data processing applications and challenges that can be addressed using NSDL approaches.

In recent years, deep learning has emerged as the dominant approach for hyperspectral image (HSI) classification. However, deep neural networks require large annotated datasets to generalize well. This limits the applicability of deep learning for real-world HSI classification problems, as manual labeling of thousands of pixels per scene is costly and time consuming. In this article, we tackle the problem of few-shot HSI classification by leveraging state-of-the-art self-supervised contrastive learning with an improved view-generation approach. Traditionally, contrastive learning algorithms heavily rely on hand-crafted data augmentations tailored for natural imagery to generate positive pairs. However, these augmentations are not directly applicable to HSIs, limiting the potential of self-supervised learning in the hyperspectral domain. To overcome this limitation, we introduce two positive pair-mining strategies for contrastive learning on HSIs. The proposed strategies mitigate the need for high-quality data augmentations, providing an effective solution for few-shot HSI classification. Through extensive experiments, we show that the proposed approach improves accuracy and label efficiency on four popular HSI classification benchmarks. Furthermore, we conduct a thorough analysis of the impact of data augmentation in contrastive learning, highlighting the advantage of our positive pair-mining approach.

Natural disasters commonly occur in all regions around the world and cause huge financial and human losses. One of the main effects of earthquakes and floods is the destruction of buildings. Photogrammetric and remote sensing (RS) data track changes and detect damages in these events. Considering the evolution in deep learning (DL) techniques, the possibility of accurate information extraction from the RS-based data is increased. DL methods effectively show the damaged regions for decision making and immediate actions for crisis management. The present study is based only on postevent RS images, which apply an encoder–decoder network composed of pretrained EfficientViTB and Yolov8 network blocks as encoder path and the modified-Unet blocks as decoder path for building damage detection (BDD). Compared with methods that use only one network in their encoder path, the presented method achieves better results. To investigate the performance of the proposed method, three datasets affected by different natural disasters are considered. The first dataset is the satellite images of the 2023 Turkey earthquake, the second dataset is associated with the satellite images of the 2023 Morocco earthquake, and the third dataset contains the satellite images of the 2023 Libya flood. The proposed method ultimately reaches the overall accuracy of 97.62%, 98.63%, and 96.43% and the kappa coefficient of 0.86, 0.85, and 0.84 for the first, second, and third dataset, respectively, which shows the proper performance of the proposed method for BDD.

In data acquisition and transmission, hyperspectral images are inevitably corrupted by additive noises, making it challenging to accurately observe and recognize the materials on the surface of the Earth. However, scholars have been addicted to developing numerous complex methods for separable two-stage denoising and anomaly detection (AD) tasks over the past years, rarely paying attention to the real effect of noises for subsequent intelligent interpretation. To this date, we propose a hierarchical integration framework for hyperspectral simultaneous AD and denoising (HyADD). Joint AD and denoising are mutually integrated and their outputs in each iteration stimulate each other, breaking through the respective performance bottlenecks of the separable two-stage scheme. Inspired by spatial–spectral gradient domain-based constraint, HyADD removes additive noises and preserves advantageous image smoothness information to improve intermediate detection performances in the iteration loop. Conversely, with the assistance of the antinoise dictionary conduction and the subspace domain-based low-rankness, the identification of anomalies with different features from the background can provide effective feedback to the denoising process. The proposed algorithm is efficiently solved by the well-designed linearized alternating direction method of multipliers with an adaptive penalty. A comparison with the existing well-known AD algorithms via simulated and real-world experiments establishes the competitiveness of the proposed HyADD with state-of-the-art methods.

Given the ever-growing availability of remote sensing data (e.g., Gaofen in China, Sentinel in the EU, and Landsat in the USA), multimodal remote sensing techniques have been garnering increasing attention and have made extraordinary progress in various Earth observation (EO)-related tasks. The data acquired by different platforms can provide diverse and complementary information. The joint exploitation of multimodal remote sensing has been proven effective in improving the existing methods of land-use/land-cover segmentation in urban environments. To boost technical breakthroughs and accelerate the development of EO applications across cities and regions, one important task is to build novel cross-city semantic segmentation models based on modern artificial intelligence technologies and emerging multimodal remote sensing data. This leads to the development of better semantic segmentation models with high transferability among different cities and regions. The Cross-City Semantic Segmentation contest is organized in conjunction with the 13th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS).

Zero-shot classification models aim to recognize image categories that are not included in the training phase by learning seen scenes with semantic information. This approach is particularly useful in remote sensing (RS) since it can identify previously unseen classes. However, most zero-shot RS scene classification approaches focus on matching visual and semantic features, while disregarding the importance of visual feature extraction, especially regarding local–global joint information. Furthermore, the visual and semantic relationships have not been thoroughly investigated due to the separate analysis of these features. To address these issues, we propose a novel zero-shot RS scene classification network based on multimodal feature interaction (MFINet). Specifically, the MFINet deploys hybrid image feature extraction networks, combining convolutional neural networks and an improved Transformer, to capture local discriminant information and long-range contextual information, respectively. Notably, we design a cross-modal feature fusion module to facilitate the MFINet, thereby enhancing relevant information in both the visual and semantic domains. Extensive experiments are conducted on the public zero-shot RS scene dataset, and the results consistently demonstrate that our proposed MFINet outperforms the state-of-the-art methods across various seen/unseen category ratios.

In recent years, remote sensing (RS) image scene classification methods have experienced notable development due to the powerful feature extraction ability of deep learning. However, current methods for RS image scene classification (RSSC) tasks struggle with handling unseen scene categories because of their reliance on large amounts of high-quality labeled data. To address this issue, zero-shot learning has been introduced into RSSC tasks, leading to the emergence of zero-shot remote sensing image scene classification (ZSRSSC) approaches. These methods refer to inference and recognition of unseen category images by comprehending seen category image features and combining category semantic information in a multimodal learning paradigm. The inference capability possessed by ZSRSSC methods can significantly improve the accuracy and applicability of RSSC in dealing with large amounts of data in the RS domain. However, this newly emerging field has produced numerous results without a comprehensive survey to organize research progress systematically. Therefore, this article systematically reviews the research progress made on these approaches, summarizes commonly used datasets and experimental setups, and compares the results of different methods. Specifically, the definition of zero-shot learning (ZSL) is reviewed first, and the mainstream development of ZSL classification methods is briefly introduced. Then, the definition of ZSRSSC is outlined, followed by an introduction to the prevalent ZSRSSC approaches and their categorization. Finally, the critical issues in these methods are analyzed, and their future development directions are discussed.

Classifiers trained on airborne hyperspectral imagery are proficient in identifying tree species in hyperdiverse tropical rainforests. However, spectral fluctuations, influenced by intrinsic and environmental factors, such as the heterogeneity of individual crown properties and atmospheric conditions, pose challenges for large-scale mapping. This study proposes an approach to assess the instability of airborne imaging spectroscopy reflectance in response to environmental variability. Through repeated overflights of two tropical forest sites in French Guiana, we explore factors that affect the spectral similarity between dates and acquisitions. By decomposing acquisitions into subsets and analyzing different sources of variability, we analyze the stability of reflectance and various vegetation indices with respect to specific sources of variability. Factors such as the variability of the viewing and sun angles or the variability of the atmospheric state shed light on the impact of sources of spectral instability, informing processing strategies. Our experiments conclude that the environmental factors that affect the canopy reflectance the most vary according to the considered spectral domain. In the short wave infrared (SWIR) domain, solar angle variation is the main source of variability, followed by atmospheric and viewing angles. In the visible and near infrared (VNIR) domain, atmospheric variability dominates, followed by solar angle and viewing angle variabilities. Despite efforts to address these variabilities, significant spectral instability persists, highlighting the need for more robust representations and improved correction methods for reliable species-specific signatures.

In PolSAR image, the presence of speckle brings significant fluctuations in the information of adjacent pixels. However, most target decomposition (TD) methods only focus on a single pixel then ignore spatial scattering information of neighborhood pixels, resulting in poor stability of decomposition features and low representation ability to terrain targets. In this paper, spatial scattering decomposition feature learning (SSDFL) model is proposed to introduce spatial information into TD. The model contains decomposition feature extraction and polarimetric data reconstruction modules. Specifically, the spatial-level and pixel-level decoders are designed to implement spatial TD and maintain the terrain details, respectively. Since the SSDFL is based on deep model learning, it can learn appropriate decomposition bases and features according to different input data. Furthermore, scattering power and component sparsity restrictions are equipped to enhance the physical scattering meanings of SSDFL model. By experiments on real PolSAR datasets, the results shown spatial information plays an irreplaceable role in TD and it also enhances the resistance to speckle. Meanwhile, the learned decomposition features could reveal the scattering characteristics of terrain targets to some extent. Finally, the effectiveness of decomposition features is verified through the features’ discriminability and the performance on real terrain classification task.

This article presents the scientific outcomes of the 2022 Hyperspectral Pansharpening Challenge organized by the 12th IEEE Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (IEEE WHISPERS 2022). The 2022 Hyperspectral Pansharpening Challenge aims at fusing a panchromatic image with hyperspectral data to get a high spatial resolution hyperspectral cube with the same spatial resolution of the panchromatic image while preserving the spectral information of hyperspectral data. Four datasets acquired by the PRISMA mission owned and managed by the Italian Space Agency have been prepared for participants. They are made available for the benefit of the scientific community. Each dataset contains a panchromatic image and a hyperspectral cube with different spatial resolutions. More than 100 registrations have been received for the event. Four teams submitted their outcomes. Since no team actually outperformed the baseline provided by the organizers, the challenge was declared inconclusive and no winner was recognized.

Deep convolutional networks are powerful for local feature learning and have shown advantages in image matching and registration. However, the significant differences between cross-modal images increase the challenge of image registration. The deep network should extract modality-invariant features to identify the matching samples and discriminative features to separate the nonmatching samples. The deep network can extract features invariant to the image modality changes by multiple nonlinear mapping layers. However, it does not inevitably lose rich details and affect the discrimination of features, degrading registration performances. This article proposes a novel deep wavelet learning network (DW-Net) for local feature learning. It incorporates spectral information into deep convolutional features for improving cross-modal image matching and registration. Specifically, this article aims to learn the multiresolution wavelet features through multilevel wavelet transform (WT) and the convolutional network. The cross-modal images are divided into low-frequency and high-frequency parts through WT. DW-Net can adaptively extract the shared features from the low-frequency part and useful details from the high-frequency part, which can enhance the modality invariance and discrimination of features. Additionally, the multiresolution wavelet features contain multiscale information and contribute to improving the matching accuracy. Extensive experiments demonstrate the significant advantages in terms of the accuracy and robustness of DW-Net on cross-modal remote sensing image registration. DW-Net can increase the image patch matching accuracy by 3.7% and improve image registration probability by 12.1%. Moreover, DW-Net shows strong generalization performances from low resolution to high resolution and from optical– synthetic aperture radar to other cross-modal image registration.

3-D building change detection (CD) methods detect more accurate multiple change maps than 2-D ones. Recent technologies, such as unmanned aerial vehicle (UAV) systems and dense image matching have made it much easier to obtain 3-D data nowadays. Developing a solution which produces an accurate map of multiple building changes, including unclassified, no building change, newly built, demolished, and taller, at an acceptable speed is a challenging issue. In this article, we address a novel 3-D building CD method based on an encoder–decoder network to detect accurate multiple changes maps automatically, in the presence of highly unbalanced remote sensing datasets. The proposed method consists of three main parts: the preprocessing and mixed augmentation (MA) step; the encoder–decoder network training; and finally the prediction step. The data are augmented by the MA method to manipulate highly unbalanced datasets. The encoder–decoder network is constructed by the Yolov7 network as the encoder path and the decoder path equipped with the convolutional layers of modified Unet). Two datasets are used in this article. The first dataset is the point clouds and orthophotos obtained from the UAV of Mashhad City in 2011 and 2016. The second dataset consists of stereo images of the GeoEye-1 satellite and the point clouds obtained from dense image matching of Tehran city in 2009 and 2013. The results show that the proposed method achieved accuracy and kappa coefficients above 94% and 0.90 for both datasets, respectively.

The accurate and fast assessment of damaged buildings following a disaster is critical for planning rescue and reconstruction efforts. The damage assessment by the traditional methods is time-consuming and with limited performance. In this article, we propose an end-to-end deep-learning network named building damage detection network-plus (BDD-Net+). The BDD-Net+ is based on a combination of convolution layers and transformer blocks. The proposed framework takes the advantage of the multiscale residual convolution blocks and self-attention layers. The proposed framework consists of four main steps: data preparation, model training, damage map generation and evaluation, and the use of an explainable artificial intelligence (XAI) framework for understanding and interpretation of the operation model. The experimental results include two representative real-world benchmark datasets (i.e., the Haiti earthquake and the Bata explosion). The obtained results illustrate that BDD-Net+ achieves excellent efficacy in comparison with other state-of-the-art methods. Furthermore, the visualization of the results by XAI shows that BDD-Net+ provides more interpretable and explainable results for damage detection than the other studied methods.

Abstract Although hyperspectral data, especially spaceborne images, are rich in spectral information, their spatial resolution is usually low due to the limitation of sensor design and other factors. Therefore, for the application of hyperspectral images, unmixing technology is a key processing technology, such as linear mixing model and its derived algorithms have made a certain progress. However, a real scene often contains both pure and mixed pixels. The existing methods usually ignore the consideration and analysis of this situation in the process of model design and simulation experiment. In this context, this paper proposes a robust linear unmixing model with the enhanced constraint of classification for hyperspectral image. In general, it designs a framework combining unmixing and classification. In the task for real scene data, endmembers are extracted first, and then the hard classification term constructed after the expansion of endmembers (training samples) based on similarity is introduced to provide the sparsity constraint of the overall model, so as to realize relatively complete adjustment and effective image unmixing under complex conditions. Considering the scene with different distributions, the simulation experiment designs several groups of data tests, including different proportions of pure and mixing pixels. The unmixing results of three simulated datasets and two real datasets show that the unmixing results of this method are better than those of the other six comparison methods. This model improves the accuracy of unmixing and realizes effective unmixing.

ABSTRACTTimely and accurate building damage mapping is essential for supporting disaster response activities. While RS satellite imagery can provide the basis for building damage map generation, detection of building damages by traditional methods is generally challenging. The traditional building damage mapping approaches focus on damage mapping based on bi-temporal pre/post-earthquake dataset extraction information from bi-temporal images, which is difficult. Furthermore, these methods require manual feature engineering for supervised learning models. To tackle the abovementioned limitation of the traditional damage detection frameworks, this research proposes a novel building damage map generation approach based only on post-event RS satellite imagery and advanced deep feature extractor layers. The proposed DL based framework is applied in an end-to-end manner without additional processing. This method can be conducted in five main steps: (1) pre-processing, (2) model training and optimization of model parameters, (3) damage mapping generation, (4) accuracy assessment, and (5) visual explanations of the proposed method’s predictions. The performance of the proposed method is evaluated by two real-world RS datasets that include Haiti-earthquake and Bata-explosion. Results of damage mapping show that the proposed method is highly efficient, yielding an OA of more than 84%, which is superior to other advanced DL-based damage detection methods.

Natural hazards can present a significant risk to road infrastructure. This infrastructure is a fundamental component of the transportation infrastructure, with significant importance. During emergencies, society heavily relies on the functionality of the road infrastructure to facilitate evacuation and access to emergency facilities. This study introduces a versatile, multi-scale framework designed to analyze accessibility within road networks during natural hazard scenarios. The first module of the framework focuses on assessing the influence of natural hazards on road infrastructure to identify damaged or blocked road segments and intersections. It relies on near real-time information, often provided by citizen science through Volunteered Geographic Information (VGI) data and Natural Language Processing (NLP) of VGI texts. The second module conducts network analysis based on freely available Open Street Map (OSM) data, differentiating between intact and degraded road networks. Four accessibility measures are employed: betweenness centrality, closeness centrality, a free-flow assumption index, and a novel alternative routing assumption measure considering congestion scenarios. The study showcases its framework through an exemplary application in California, the United States, considering different hazard scenarios, where degraded roads and connected roads impacted by the hazard can be identified. The road extraction methodology allows the extraction of 75% to 100% of the impacted roads mentioned in VGI text messages for the respective case studies. In addition to the directly extracted impacted roads, constructing the degraded network also involves finding road segments that overlap with hazard impact zones, as these are at risk of being impacted. Conducting the network analysis with the four different measures on the intact and degraded network, changes in network accessibility due to the impacts of hazards can be identified. The results show that using each measure is justified, as each measure could demonstrate the accessibility change. However, their combination and comparison provide valuable insights. In conclusion, this study successfully addresses the challenges of developing a generic, complete framework from impact extraction to network analysis independently of the scale and characteristics of road network types.

Hyperspectral images (HSIs) are widely used to identify and characterize objects in scenes of interest, but they are associated with high acquisition costs and low spatial resolutions. With the development of deep learning, HSI reconstruction from low-cost and high-spatial-resolution RGB images has attracted widespread attention. It is an inexpensive way to obtain HSIs via the spectral reconstruction (SR) of RGB data. However, due to a lack of consideration of outdoor solar illumination variation in existing reconstruction methods, the accuracy of outdoor SR remains limited. In this paper, we present an attention neural network based on an adaptive weighted attention network (AWAN), which considers outdoor solar illumination variation by prior illumination information being introduced into the network through a basic 2D block. To verify our network, we conduct experiments on our Variational Illumination Hyperspectral (VIHS) dataset, which is composed of natural HSIs and corresponding RGB and illumination data. The raw HSIs are taken on a portable HS camera, and RGB images are resampled directly from the corresponding HSIs, which are not affected by illumination under CIE-1964 Standard Illuminant. Illumination data are acquired with an outdoor illumination measuring device (IMD). Compared to other methods and the reconstructed results not considering solar illumination variation, our reconstruction results have higher accuracy and perform well in similarity evaluations and classifications using supervised and unsupervised methods.

Natural hazards such as wildfires have proven to be more frequent in recent years, and to minimize losses and activate emergency response, it is necessary to estimate their impact quickly and consequently identify the most affected areas. Volunteered geographic information (VGI) data, particularly from the social media platform Twitter, now X, are emerging as an accessible and near-real-time geoinformation data source about natural hazards. Our study seeks to analyze and evaluate the feasibility and limitations of using tweets in our proposed method for fire area assessment in near-real time. The methodology involves weighted barycenter calculation from tweet locations and estimating the affected area through various approaches based on data within tweet texts, including viewing angle to the fire, road segment blocking information, and distance to fire information. Case study scenarios are examined, revealing that the estimated areas align closely with fire hazard areas compared to remote sensing (RS) estimated fire areas, used as pseudo-references. The approach demonstrates reasonable accuracy with estimation areas differing by distances of 2 to 6 km between VGI and pseudo-reference centers and barycenters differing by distances of 5 km on average from pseudo-reference centers. Thus, geospatial analysis on VGI, mainly from Twitter, allows for a rapid and approximate assessment of affected areas. This capability enables emergency responders to coordinate operations and allocate resources efficiently during natural hazards.

Due to advances in remote sensing satellite imaging and image processing technologies and their wide applications, intelligent remote sensing satellites are facing an opportunity for rapid development. The key technologies, standards, and laws of intelligent remote sensing satellites are also experiencing a series of new challenges. Novel concepts and key technologies in the intelligent hyperspectral remote sensing satellite system have been proposed since 2011. The aim of these intelligent remote sensing satellites is to provide real-time, accurate, and personalized remote sensing information services. This article reviews the current developments in new-generation intelligent remote sensing satellite systems, with a focus on intelligent remote sensing satellite platforms, imaging payloads, onboard processing systems, and other key technological chains. The technological breakthroughs and current defects of intelligence-oriented designs are also analyzed. Intelligent remote sensing satellites collect personalized remote sensing data and information, with real-time data features and information interaction between remote sensing satellites or between satellites and the ground. Such developments will expand the use of remote sensing applications beyond government departments and industrial users to a massive number of individual users. However, this extension faces challenges regarding privacy protection, societal values, and laws regarding the sharing and distribution of data and information.

Recently, the semi-supervised graph convolutional network (SSGCN) has been verified effective for hyperspectral image (HSI) classification. However, constrained by the limited training data and spectral uncertainty, the classification performance is remained to be further improved. Moreover, attribute to the massive data, the SSGCN with complex computation is generally too time- and resource-consuming to be applicable in real-time needs. To conquer these issues, we propose an efficient symmetric graph metric learning (SGML) framework by incorporating metric learning into the SSGCN paradigm. Specifically, we first conduct multilevel pixel-to-superpixel projection (P-SP) on the HSI to investigate the multiscale spatial information, where the suitable superpixel numbers are adaptively determined. Then, to extract more expressive representations, we design a new structure denoted as GSvolution, comprising the graph convolution (G-Conv) and a novel self-channel-enhanced convolution (S-Conv), to propagate the labeled and unlabeled graph node information and simultaneously enhance the critical intranode channel features. Finally, the superpixel node features are reprojected to the pixel level (SP-P) so that the distilled multistream features can be integrated to obtain the final decision. Noticeably, this ingenious symmetric mechanism (P-SP and SP-P) can alleviate the spectral variability and facilitate the framework to be an efficient model. Furthermore, in the metric learning module, we propose an innovative metric loss function to enhance the discrimination of the embedding features, i.e., inter class far apart and intraclass close. In the experiments, we demonstrate that the classification capacity of the proposed SGML can surpass the comparators on three benchmark data sets.

Hyperspectral imaging (HSI) has been extensively utilized in many real-life applications because it benefits from the detailed spectral information contained in each pixel. Notably, the complex characteristics, i.e., the nonlinear relation among the captured spectral information and the corresponding object of HSI data, make accurate classification challenging for traditional methods. In the last few years, deep learning (DL) has been substantiated as a powerful feature extractor that effectively addresses the nonlinear problems that appeared in a number of computer vision tasks. This prompts the deployment of DL for HSI classification (HSIC) which revealed good performance. This survey enlists a systematic overview of DL for HSIC and compared state-of-the-art strategies of the said topic. Primarily, we will encapsulate the main challenges of TML for HSIC and then we will acquaint the superiority of DL to address these problems. This article breaks down the state-of-the-art DL frameworks into spectral-features, spatial-features, and together spatial–spectral features to systematically analyze the achievements (future research directions as well) of these frameworks for HSIC. Moreover, we will consider the fact that DL requires a large number of labeled training examples whereas acquiring such a number for HSIC is challenging in terms of time and cost. Therefore, this survey discusses some strategies to improve the generalization performance of DL strategies which can provide some future guidelines.

Wildfires are considered as one of the most disturbing factors in forest areas and high-density vegetation regions. The mapping of wildfires is particularly important for fire prediction and burned biomass estimation. Therefore, accurate and timely mapping of burned areas is of great importance and has a key role in disaster management. Estimation of burned areas from multispectral datasets is challenging, because of the complexity of the background and the different reflections of wildfires on the targets. To this end, this study presents a novel burned-area mapping framework based on the fusion of multitemporal Sentinel-1 coherence imagery and post-event Sentinel-2 imagery. The proposed framework uses hybrid quadratic morphological (QM) operations and convolution layers for deep feature extraction. The proposed architecture is known as the QMDNN-Net, where the overall framework of QMDNN-Net is based on a deep Siamese network. QMDNN-Net has double streams for extracting deep features from multitemporal coherence data and Sentinel-2 imagery. The streams are similar to each other and have the same number of group-dilated convolution blocks and QM layers. QMDNN-Net is defined based on quadratic dilation and erosion, and then it takes the average of these as output. The results of wildfire mapping with QMDNN-Net are evaluated here with two real-world wildfire datasets based on Sentinel-1 and Sentinel-2 imagery. The results show that QMDNN-Net achieves an overall accuracy and Kappa coefficient of 95.5% and 0.9, respectively, and outperforms other state-of-the-art methods.

Hyperspectral imagery (HSI) classification, which attempts to assign hyperspectral pixels with proper labels, has drawn significant attention in various applications. Recently, the graph-based semisupervised learning (SSL) approaches have shown the outstanding ability to handle the situation of limited labeled data for classification task. However, it is hard to construct the pairwise adjacent graph due to the high dimensionality of hyperspectral data. Besides, the graph-based SSL models are usually decided by a single classifier, which fail to effectively learn the complex structures and intrinsic properties of HSI. To address these problems, we propose a novel graph-based SSL classification model for HSI, which is based on random multigraphs construction and ensemble strategy (RMGE). Specifically, the anchor graph (AG) is constructed with spatial–spectral features, which integrates the spatial characteristics extracted by local binary pattern on each selected spectrum, preserving fine structures of local region. In order to enhance the discriminative capability of the classifier and avoid the trivial solution, the maximum entropy regularization is added into adjacent AG model. In addition, to capture the diversity of HSI data effectively, we design the ensemble framework by employing multiple AGs to learn HSI features. Experiments conducted on real hyperspectral datasets indicate that the proposed RMGE shows better performance than that of state-of-the-art approaches.

With the extremely rapid advances in remote sensing (RS) technology, a great quantity of Earth observation (EO) data featuring considerable and complicated heterogeneity are readily available nowadays, which renders researchers an opportunity to tackle current geoscience applications in a fresh way. With the joint utilization of EO data, much research on multimodal RS data fusion has made tremendous progress in recent years, yet these developed traditional algorithms inevitably meet the performance bottleneck due to the lack of the ability to comprehensively analyze and interpret strongly heterogeneous data. Hence, this non-negligible limitation further arouses an intense demand for an alternative tool with powerful processing competence. Deep learning (DL), as a cutting-edge technology, has witnessed remarkable breakthroughs in numerous computer vision tasks owing to its impressive ability in data representation and reconstruction. Naturally, it has been successfully applied to the field of multimodal RS data fusion, yielding great improvement compared with traditional methods. This survey aims to present a systematic overview in DL-based multimodal RS data fusion. More specifically, some essential knowledge about this topic is first given. Subsequently, a literature survey is conducted to analyze the trends of this field. Some prevalent sub-fields in the multimodal RS data fusion are then reviewed in terms of the to-be-fused data modalities, i.e., spatiospectral, spatiotemporal, light detection and ranging-optical, synthetic aperture radar-optical, and RS-Geospatial Big Data fusion. Furthermore, We collect and summarize some valuable resources for the sake of the development in multimodal RS data fusion. Finally, the remaining challenges and potential future directions are highlighted.

Land cover classification (LCC) is an important application in remote sensing data interpretation and invariably faces big intra-class variance and sample imbalance in remote sensing images. The optical image is obtained by satellites capturing the spectral information of the Earth’s surface, and the synthetic aperture radar (SAR) image is produced by the satellite actively transmitting and receiving the electromagnetic wave signals reflected from land covers. Because of the limitations of the optical image, a single modality (optical image) might be disturbed by external conditions, especially complex weather. Using heterogeneous SAR and optical images for LCC can reduce the negative impact caused by single-modal data damage, and multi-modal data can also be used as supplementary information to enhance classification accuracy. However, general LCC methods mainly focus on remote sensing data of a single modality without fully considering the multi-modalities of land covers. Therefore, we propose a dual-stream deep high-resolution network (DDHRNet) to deeply integrate SAR and optical data at the feature level in every branch. The network can effectively exploit the complementary information in heterogeneous images. It improves classification performance and achieves significant improvements in the classification of clouded images. A multi-modal squeeze-and-excitation (MSE) module is also utilized to fuse the features. Compared with the ordinary methods, MSE modules can lead to an improvement of about 1% to 5% in overall accuracy (OA), Kappa coefficients, and mean intersection over union (mIoU). Besides, in order to evaluate our method, we describe in detail the preprocessing process of Gaofen-2 (GF2) and Gaofen-3 (GF3) data before they are used in the LCC task. The experiments show that the proposed method performs well compared with other excellent segmentation methods and obtains the best performance on heterogeneous images from GF2 and GF3. The code and datasets are available at: https://github.com/XD-MG/DDHRNet.

Accurate and timely mapping of wildfire burned areas is crucial for post-fire management, planning, and next subsequent actions. The monitoring and mapping of the burned area by traditional and common methods are time-consuming and challenging while is vital to propose an advanced burned area detection framework for achieving reliable results. To this end, this study proposed a novel End-to-End framework based on deep learning and post-fire Sentinel-2 imagery. The proposed framework known as Burnt-Net combines quadratic morphological operators and standard convolution layers. The multi-patch multi-level residual morphological (MP-MRM) blocks are the main part of the decoder part of the Burnt-Net while the encoder part uses the multi-level residual morphological and transpose convolution layers. To evaluate the efficiency of Burnt-Net the post-fire Sentinel-2 for the latest wildfires over different countries was collected and then, the model was trained and evaluated based on them. Furthermore, the most common deep learning-based model implemented for comparing the result of burned areas by the proposed Burnt-Net. The results of burned areas mapping show the Burnt-Net is robust in the detection of burned areas and provides a mean accuracy of more than 97% by overall accuracy (OA). Furthermore, the Burnt-Net is fast and can provide the burned area map in the near real-time.

Over the past decade, Convolutional Networks (ConvNets) have renewed the perspectives of the research and industrial communities. Although this deep learning technique may be composed of multiple layers, its core operation is the convolution, an important linear filtering process. Easy and fast to implement, convolutions actually play a major role, not only in ConvNets, but in digital image processing and analysis as a whole, being effective for several tasks. However, aside from convolutions, researchers also proposed and developed non-linear filters, such as operators provided by mathematical morphology. Even though these are not so computationally efficient as the linear filters, in general, they are able to capture different patterns and tackle distinct problems when compared to the convolutions. In this paper, we propose a new paradigm for deep networks where convolutions are replaced by non-linear morphological filters. Aside from performing the operation, the proposed Deep Morphological Network (DeepMorphNet) is also able to learn the morphological filters (and consequently the features) based on the input data. While this process raises challenging issues regarding training and actual implementation, the proposed DeepMorphNet proves to be able to extract features and solve problems that traditional architectures with standard convolution filters cannot.

In this article, we elaborate on the scientific outcomes of the 2021 Data Fusion Contest (DFC2021), which was organized by the Image Analysis and Data Fusion Technical Committee of the IEEE Geoscience and Remote Sensing Society, on the subject of geospatial artificial intelligence for social good. The ultimate objective of the contest was to model the state and changes of artificial and natural environments from multimodal and multitemporal remotely sensed data towards sustainable developments. DFC2021 consisted of two challenge tracks: Detection of settlements without electricity (DSE) and multitemporal semantic change detection. We focus here on the outcome of the DSE track. This article presents the corresponding approaches and reports the results of the best-performing methods during the contest.

Floods are one of the most destructive natural disasters, causing financial and human losses every year. As a result, reliable Flood Susceptibility Mapping (FSM) is required for effective flood management and reducing its harmful effects. In this study, a new machine learning model based on the Cascade Forest Model (CFM) was developed for FSM. Satellite imagery, historical reports, and field data were used to determine flood-inundated areas. The database included 21 flood-conditioning factors obtained from different sources. The performance of the proposed CFM was evaluated over two study areas, and the results were compared with those of other six machine learning methods, including Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), Deep Neural Network (DNN), Light Gradient Boosting Machine (LightGBM), Extreme Gradient Boosting (XGBoost), and Categorical Boosting (CatBoost). The result showed CFM produced the highest accuracy compared to other models over both study areas. The Overall Accuracy (AC), Kappa Coefficient (KC), and Area Under the Receiver Operating Characteristic Curve (AUC) of the proposed model were more than 95%, 0.8, 0.95, respectively. Most of these models recognized the southwestern part of the Karun basin, northern and northwestern regions of the Gorganrud basin as susceptible areas.

Pan-sharpening, which fuses the high-resolution panchromatic (PAN) image and the low-resolution multispectral image (MSI), is a hot topic in remote sensing. Recently, deep learning technology has been successfully applied in pan-sharpening. However, the existing methods ignore that the MSI and PAN image are at different resolutions and use the same networks to extract features of the two images. To address this problem, we propose a two-stream deep learning architecture, called coupled multiscale convolutional neural network, for pan-sharpening. The proposed network has three components, feature extraction subnetworks, fusion layer, and super-resolution subnetwork. In the feature extraction subnetworks, two subnetworks are used to extract the features of the MSI and PAN image separately. Different sizes of convolutional kernels are used in the first layers due to the different spatial resolutions. Thus, the source images are mapped to the similar scale. Then a multiscale asymmetric convolution factorization is used to extract features at different scales. In the fusion layer, the two feature extraction subnetworks are coupled. Features at the same scale are first summed, and then the features of all scales are concatenated as one feature map. At last, a super-resolution subnetwork is used to generate the high-resolution MSI. Experimental results on both synthetic and real data sets demonstrate that the proposed method outperforms the other state-of-the-art pan-sharpening methods.

Hyperspectral (HS) pansharpening aims at fusing a low-resolution HS (LRHS) image with a panchromatic image to obtain a full-resolution HS image. Most of the existing HS pansharpening approaches are usually based on traditional multispectral pansharpening techniques, which are not especially tailored for two inherent challenges of the HS pansharpening, i.e., much wider spectral range gap between the two kinds of images and having to recover details in many continuous spectral bands simultaneously. In this article, we develop new spectral-fidelity convolutional neural networks (called HSpeNets) for HS pansharpening to keep the fidelity of a pansharpened image to its true spectra as much as possible. Our methods particularly focus on the decomposability of HS details, accordingly synthesizing these details progressively, and meanwhile introduce a spectral-fidelity loss. We give theoretical justifications and provide detailed experimental results, showing the superiorities of the proposed HSpeNets with regard to other state-of-the-art pansharpening approaches.

The large-scale variation issue in high-resolution aerial images significantly lowers the accuracy of segmenting small objects. For a deep-learning-based semantic segmentation model, the main reason is that the deeper layers generate high-level semantics over considerably large receptive fields, thus improving the accuracy for large objects but ignoring small objects. Although the low-level features extracted by shallow layers contain small-object information, large-object information has predominant effects. When the model, using low-level features, is trained, the large objects push the small objects aside. This observation motivates us to propose a novel reverse difference mechanism (RDM). The RDM eliminates the predominant effects of large objects and highlights small objects from low-level features. Based on the RDM, a novel semantic segmentation method called the reverse difference network (RDNet) is designed. In the RDNet, a detailed stream is proposed to produce small-object semantics by enhancing the output of RDM. A contextual stream for generating high-level semantics is designed by fully accumulating contextual information to ensure the accuracy of the segmentation of large objects. Both high-level and small-object semantics are concatenated when the RDNet performs predictions. Thus, both small- and large-object information is depicted well. Two semantic segmentation benchmarks containing vital small objects are used to fully evaluate the performance of the RDNet. Compared with existing methods that exhibit good performance in segmenting small objects, the RDNet has lower computational complexity and achieves 3.9–18.9% higher accuracy in segmenting small objects.

Documenting the inter-annual variability and the long-term trend of the glacier snow line altitude is highly relevant to document the evolution of glacier mass changes. Automatically identifying the snow line on glaciers is challenging; recent developments in machine learning approaches show promise to tackle this issue. This manuscript presents a proof of concept of machine learning approaches applied to multi-spectral images to detect the snow line and quantify its average altitude. The tested approaches include the combination of different image processing and classification methods, and takes into account cast shadows. The efficiency of these approaches is evaluated on mountain glaciers in the European Alps by comparing the results with manually annotated data. Solutions provided by the different approaches are robust when compared to the ground truth’s snow lines, with a Pearson’s correlation ranging from 79% to 96% depending on the method. However, the tested approaches may fail when snow lines are not continuous or exhibit a strong change of elevation. The major advantage over the state of the art is that the proposed approach does not require one calibration per glacier.

Polarimetric synthetic aperture radar (PolSAR) data can be acquired at all times and are not impacted by weather conditions. They can efficiently capture geometrical and geographical structures on the ground. However, due to the complexity of the data and the difficulty of data availability, PolSAR image scene classification remains a challenging task. To this end, in this paper, a low-rank constrained multimodal tensor representation method (LR-MTR) is proposed to integrate PolSAR data in multimodal representations. To preserve the multimodal polarimetric information simultaneously, the target decompositions in a scene from multiple spaces (e.g., Freeman, H/A/α, Pauli, etc.) are exploited to provide multiple pseudo-color images. Furthermore, a representation tensor is constructed via the representation matrices and constrained by the low-rank norm to keep the cross-information from multiple spaces. A projection matrix is also calculated by minimizing the differences between the whole cascaded data set and the features in the corresponding space. It also reduces the redundancy of those multiple spaces and solves the out-of-sample problem in the large-scale data set. To support the experiments, two new PolSAR image data sets are built via ALOS-2 full polarization data, covering the areas of Shanghai, China, and Tokyo, Japan. Compared with state-of-the-art (SOTA) dimension reduction algorithms, the proposed method achieves the best quantitative performance and demonstrates superiority in fusing multimodal PolSAR features for image scene classification.

Hyperspectral image-anomaly detection (HSI-AD) has become one of the research hotspots in the field of remote sensing. Because HSI’s features of integrating image and spectrum provide a considerable data basis for abnormal object detection, HSI-AD has a huge application potential in HSI analysis. It is difficult to effectively extract a large number of nonlinear features contained in HSI data using traditional machine learning methods, and deep learning has incomparable advantages in the extraction of nonlinear features. Therefore, deep learning has been widely used in HSI-AD and has shown excellent performance. This review systematically summarizes the related reference of HSI-AD based on deep learning and classifies the corresponding methods into performance comparisons. Specifically, we first introduce the characteristics of HSI-AD and the challenges faced by traditional methods and introduce the advantages of deep learning in dealing with these problems. Then, we systematically review and classify the corresponding methods of HSI-AD. Finally, the performance of the HSI-AD method based on deep learning is compared on several mainstream data sets, and the existing challenges are summarized. The main purpose of this article is to give a more comprehensive overview of the HSI-AD method to provide a reference for future research work.

Wildfires are one of the most destructive natural disasters that can affect our environment, with significant effects also on wildlife. Recently, climate change and human activities have resulted in higher frequencies of wildfires throughout the world. Timely and accurate detection of the burned areas can help to make decisions for their management. Remote sensing satellite imagery can have a key role in mapping burned areas due to its wide coverage, high-resolution data collection, and low capture times. However, although many studies have reported on burned area mapping based on remote sensing imagery in recent decades, accurate burned area mapping remains a major challenge due to the complexity of the background and the diversity of the burned areas. This paper presents a novel framework for burned area mapping based on Deep Siamese Morphological Neural Network (DSMNN-Net) and heterogeneous datasets. The DSMNN-Net framework is based on change detection through proposing a pre/post-fire method that is compatible with heterogeneous remote sensing datasets. The proposed network combines multiscale convolution layers and morphological layers (erosion and dilation) to generate deep features. To evaluate the performance of the method proposed here, two case study areas in Australian forests were selected. The framework used can better detect burned areas compared to other state-of-the-art burned area mapping procedures, with a performance of >98% for overall accuracy index, and a kappa coefficient of >0.9, using multispectral Sentinel-2 and hyperspectral PRISMA image datasets. The analyses of the two datasets illustrate that the DSMNN-Net is sufficiently valid and robust for burned area mapping, and especially for complex areas.

Deep learning has been widely used to fuse multi-sensor data for classification. However, current deep learning architecture for multi-sensor data fusion might not always perform better than single data source, especially for the fusion of hyperspectral and light detection and ranging (LiDAR) remote sensing data for tree species mapping in complex, closed forest canopies. In this paper, we propose a new deep fusion framework to integrate the complementary information from hyperspectral and LiDAR data for tree species mapping. We also investigate the fusion of either “single-band" or multi-band (i.e., fullwaveform) LiDAR with hyperspectral data for tree species mapping. Additionally, we provide a solution to estimate the crown size of tree species by the fusion of multi-sensor data. Experimental results on fusing real APEX hyperspectral and LiDAR data demonstrate the effectiveness of the proposed deep fusion framework. Compared to using only single data source or current deep fusion architecture, our proposed method yields improvements in overall and average classification accuracies ranging from 82.21% to 87.10% and 76.71% to 83.45%, respectively.

Managing transmission and storage of hyperspectral (HS) images can be extremely difficult. Thus, the dimensionality reduction of HS data becomes necessary. Among several dimensionality reduction techniques, transform-based have found to be effective for HS data. While spatial transformation techniques provide good compression rates, the choice of the spectral decorrelation approaches can have strong impact on the quality of the compressed image. Since HS images are highly correlated within each spectral band and in particular across neighboring bands, the choice of a decorrelation method allowing to retain as much information content as possible is desirable. From this point of view, several methods based on PCA and Wavelet have been presented in the literature. In this paper, we propose the use of NLPCA transform as a method to reduce the spectral dimensionality of HS data. NLPCA represents in a lower dimensional space the same information content with less features than PCA. In these terms, aim of this research is focused on the analysis of the results obtained through the spectral decorrelation phase rather than taking advantage of both spectral and spatial compression. Experimental results assessing the advantage of NLPCA with respect to standard PCA are presented on four different HS datasets.

Large-scale land-cover classification using a supervised algorithm is a challenging task. Enormous efforts have been made to manually process and check the production of national land-cover maps. This has led to complex pre- and post-processing and even the production of inaccurate mapping products from large-scale remote sensing images. Inspired by the recent success of deep learning techniques, in this study we provided a feasible automatic solution for improving the quality of national land-cover maps. However, the application of deep learning to national land-cover mapping remains limited because only small-scale noisy labels are available. To this end, a mutual transfer network MTNet was developed. MTNet is capable of learning better feature representations by mutually transferring pre-trained models from time-series of data and fine-tuning current data. An interactive training strategy such as this can effectively alleviate the effects of inaccurate or noisy labels and unbalanced sample distributions, thus yielding a relatively stable classification system. Extensive experiments were conducted by focusing on several representative regions to evaluate the classification results of our proposed method. Quantitative results showed that the proposed MTNet outperformed its baseline model about 1%, and the accuracy can be improved up to 6.45% compared with the model trained by the training set of another year. We also visualized the national classification maps generated by MTNet for two different time periods to quantitatively analyze the performance gain. It was concluded that the proposed MTNet provides an efficient method for large-scale land cover mapping.

Hyperspectral analysis is a well-established technique that can be suitably implemented in several application fields, including materials science. This approach allows us to deal with data samples containing spatial and spectral information at very high resolution, thus enabling us to evaluate materials properties at a nanoscale level. As a proof of concept, hyperspectral imaging was here considered to investigate 3D printed polymer matrix composites, considering graphene oxide (GO) as a nanofiller. Commercial polycaprolactone and polylactic acid filaments were firstly treated with GO to be then printed into testing specimens. Raman analysis was performed to assess the GO distribution on samples surface by mapping different regions of interest and the collected data were the input of a custom-made algorithm for hyperspectral image analysis, tailored to detect the GO signature. Findings showed a valuable matching to Raman maps and were also characterized by the positive feature of avoiding to set specific conditions to perform the investigation as GO Raman distribution was carried out by fixing the wavenumber at 1580 cm−1, which is representative of the G band of the nanofiller. This occurrence might lead to an uneven intensity representation related to possible peak shifts which can bias the acquired results. Differently, hyperspectral imaging needs a minimal set of data input, i.e., the spectral signatures of neat materials, to directly identify the searched nanomaterial. More in-depth investigations need to be performed to fully validate the proposed approach, but the here presented results already show the potential and versatility of hyperspectral analysis in the materials science field.

In recent years, fusing hyperspectral images (HSIs) and multispectral images (MSIs) to acquire super-resolution images (SRIs) has been in the spotlight and gained tremendous attention. However, some current methods, such as those based on low rank matrix decomposition, also have a fair share of challenges. These algorithms carry out the matrixing process for the original image tensor, which will lose the structure information of the original image. In addition, there is no corresponding theory to prove whether the algorithm can guarantee the accurate restoration of the fused image due to the non-uniqueness of matrix decomposition. Moreover, degenerate operators are usually unknown or difficult to estimate in some practical applications. In this paper, an image fusion method based on joint tensor decomposition (JTF) is proposed, which is more effective and more applicable to the circumstance that degenerate operators are unknown or tough to gauge. Specifically, in the proposed JTF method, we consider SRI as a three-dimensional tensor and redefine the fusion problem with the decomposition issue of joint tensors. We then formulate the JTF algorithm, and the experimental results certify the superior performance of the proposed method in comparison to the current popular schemes.

The complementary nature of different modalities and multiple bands used in remote sensing data is helpful for tasks such as change detection and the prediction of agricultural variables. Nonetheless, correctly processing a multi-modal dataset is not a simple task, owing to the presence of different data resolutions and formats. In the past few years, graph-based methods have proven to be a useful tool in capturing inherent data similarity, in spite of different data formats, and preserving relevant topological and geometric information. In this paper, we propose a graph-based data fusion algorithm for remotely sensed images applied to (i) data-driven semi-unsupervised change detection and (ii) biomass estimation in rice crops. In order to detect the change, we evaluated the performance of four competing algorithms on fourteen datasets. To estimate biomass in rice crops, we compared our proposal in terms of root mean squared error (RMSE) concerning a recent approach based on vegetation indices as features. The results confirm that the proposed graph-based data fusion algorithm outperforms state-of-the-art methods for change detection and biomass estimation in rice crops.

This paper proposes three feature extraction (FE) methods based on density estimation for hyperspectral images (HSIs). The methods are a mixture of factor analyzers (MFA), deep MFA (DMFA), and supervised MFA (SMFA). The MFA extends the Gaussian mixture model to allow a low-dimensionality representation of the Gaussians. DMFA is a deep version of MFA and consists of a two-layer MFA, i.e, samples from the posterior distribution at the first layer are input to an MFA model at the second layer. SMFA consists of single-layer MFA and exploits labeled information to extract features of HSI effectively. Based on these three FE methods, the paper also proposes a framework that automatically extracts the most important features for classification from an HSI. The overall accuracy of a classifier is used to automatically choose the optimal number of features and hence performs dimensionality reduction (DR) before HSI classification. The performance of MFA, DMFA, and SMFA FE methods are evaluated and compared to five different types of unsupervised and supervised FE methods by using four real HSIs datasets.

Le démélange d'images hyperspectrales vise à extraire les spectres des matériaux purs de la scène observée (pôles de mélange ou endmembers), ainsi que leurs proportions dans chaque pixel de l'image (abondances). La plupart des algorithmes s'appuient sur des techniques classiques d'extraction d'endmembers qui peuvent échouer dans des scénarios difficiles. Dans cet article, nous abordons ce problème en même temps que la variabilité des matériaux en considérant qu'un endmember est une direction dans l'espace ambiant plutôt qu'un seul point. Sous ce paradigme, nous proposons un algorithme pour fournir des spectres de référence robustes. Nous montrons le potentiel de l'algorithme proposé sur un jeu de données synthétiques en utilisant des spectres réels présentant de la variabilité intra-classe, et une image d'une scène naturelle complexe.

Concerning the strengths and limitations of multispectral and airborne LiDAR data, the fusion of such datasets can compensate for the weakness of each other. This work have investigated the integration of multispectral and airborne LiDAR data for the land cover mapping of large urban area. Different LiDAR-derived features are involoved, including height, intensity, and multiple-return features. However, there is limited knowledge relating to the integration of multispectral and LiDAR data including three feature types for the classification task. Furthermore, a little contribution has been devoted to the relative importance of input features and the impact on the classification uncertainty by using multispectral and LiDAR. The key goal of this study is to explore the potenial improvement by using both multispectral and LiDAR data and to evaluate the importance and uncertainty of input features. Experimental results revealed that using the LiDAR-derived height features produced the lowest classification accuracy (83.17%). The addition of intensity information increased the map accuracy by 3.92 percentage points. The accuracy was further improved to 87.69% with the addition multiple-return features. A SPOT-5 image produced an overall classification accuracy of 86.51%. Combining spectral and spatial features increased the map accuracy by 6.03 percentage points. The best result (94.59%) was obtained by the combination of SPOT-5 and LiDAR data using all available input variables. Analysis of feature relevance demonstrated that the normalized digital surface model (nDSM) was the most beneficial feature in the classification of land cover. LiDAR-derived height features were more conducive to the classification of urban area as compared to LiDAR-derived intensity and multiple-return features. Selecting only 10 most important features can result in higher overall classification accuracy than all scenarios of input variables except the feature of entry scenario using all available input features. The variable importance varied a very large extent in the light of feature importance per land cover class. Results of classification uncertainty suggested that feature combination can tend to decrease classification uncertainty for different land cover classes, but there is no “one-feature-combination-fits-all” solution. The values of classification uncertainty exhibited significant differences between the land cover classes, and extremely low uncertainties were revealed for the water class. However, it should be noted that using all input variables resulted in relatively lower classification uncertainty values for most of the classes when compared to other input features scenarios.

We propose to replace traditional spectral index methods by unsupervised spectral unmixing methods for the exploration of large datasets of planetary hyperspectral images. The main goal of this article is to test the ability of these analysis techniques to automatically extract the spectral signatures of the species present on the surface and to map their abundances accurately and with an acceptable processing time. We consider observations of the surface of Mars acquired by the imaging spectrometer OMEGA aboard MEX as a case study. The moderate spatial resolution (≈300 m/pixel at best) of this instrument implies the systematic existence of geographical mixtures possibly conjugated with non-linear (e.g., intimate) mixtures. We examine the sensitivity of a series of state-of-the-art methods of unmixing to the intrinsic spectral variability of the species in the image and to intimate assemblages of compounds. This study is made possible thanks to the use of well-controlled synthetic data and a real OMEGA image, for which the present icy species (water and carbon dioxide ices) and their characteristic spectra are widely known by the planetary community. Furthermore, reference maps of component abundances are built by the inversion of a more realistic physical model (simulating the propagation of solar light through the atmosphere and reflected back to the sensor) in order to validate the methods with the real image by comparison with the maps extracted by unmixing. The results produced by the processing pipeline of the eigenvalue likelihood maximization (ELM), vertex component analysis (VCA) and non-negativity condition least squares error estimators (NNLS) are the most robust to non-linear effects, highly-mixed pixels and different types of mixtures. Despite this fact, the produced results are not always the best because the VCA method assumes the existence of pure pixels in the image, that is pixels completely occupied by a single species. However, this pipeline is very fast and provides endmember spectra that are always interpretable. Finally, it produces more accurate distribution maps than the spectral index methods. More generally, the potential benefits of unsupervised spectral unmixing methods in planetary exploration is emphasized.

The characterization of snow extent is critical for a wide range of applications. Since 1966, snow maps at different spatial resolutions have been produced using various satellite sensor images. Nowadays, the most widely used products are likely those derived from Moderate-Resolution Imaging Spectroradiometer (MODIS) data, which cover the whole Earth at a near-daily frequency. There are a variety of snow mapping methods for MODIS data, based on different methodologies and applied at different spatial resolutions. Up to now, all these products have been tested and evaluated separately. This study aims to compare the methods currently available for retrieving snow from MODIS data. The focus is on fractional snow cover, which represents the snow cover area at the subpixel level. We examine the two main approaches available for generating such products from MODIS data; namely, linear regression of the Normalized Difference Snow Index (NDSI) and spectral unmixing (SU). These two approaches have resulted in several methods, such as MOD10A1 (the NSIDC MODIS snow product) for NDSI regression, and MODImLAB for SU. The assessment of these approaches was carried out using higher resolution binary snow maps (i.e., showing the presence or absence of snow) at spatial resolutions of 10, 20, and 30 m, produced by SPOT 4, SPOT 5, and LANDSAT-8, respectively. Three areas were selected in order to provide landscape diversity: the French Alps (117 dates), the Pyrenees (30 dates), and the Moroccan Atlas (24 dates). This study investigates the impact of reference maps on accuracy assessments, and it is suggested that NDSI-based high spatial resolution reference maps advantage NDSI medium-resolution snow maps. For MODIS snow maps, the results show that applying an NDSI approach to accurate surface reflectance corrected for topographic and atmospheric effects generally outperforms other methods for the global retrieval of snow cover area. The improvements to the newer version of MOD10A1 (Collection 6) compared to the older version (Collection 5) are significant. Products based on SU provide a good alternative and more accurate retrieval of the snow fraction where wider ranges of land covers are concerned. The fusion process and its resulting 250 m spatial resolution product improve snow line retrieval. False detection in mixed pixels, probably due to the spectral variability associated with the various materials in the spectral mixture, has been identified as an area that will require improvement.

Hyperspectral remote sensing is based on measuring the scattered and reflected electromagnetic signals from the Earth’s surface emitted by the Sun. The received radiance at the sensor is usually degraded by atmospheric effects and instrumental (sensor) noises which include thermal (Johnson) noise, quantization noise, and shot (photon) noise. Noise reduction is often considered as a preprocessing step for hyperspectral imagery. In the past decade, hyperspectral noise reduction techniques have evolved substantially from two dimensional bandwise techniques to three dimensional ones, and varieties of low-rank methods have been forwarded to improve the signal to noise ratio of the observed data. Despite all the developments and advances, there is a lack of a comprehensive overview of these techniques and their impact on hyperspectral imagery applications. In this paper, we address the following two main issues; (1) Providing an overview of the techniques developed in the past decade for hyperspectral image noise reduction; (2) Discussing the performance of these techniques by applying them as a preprocessing step to improve a hyperspectral image analysis task, i.e., classification. Additionally, this paper discusses about the hyperspectral image modeling and denoising challenges. Furthermore, different noise types that exist in hyperspectral images have been described. The denoising experiments have confirmed the advantages of the use of low-rank denoising techniques compared to the other denoising techniques in terms of signal to noise ratio and spectral angle distance. In the classification experiments, classification accuracies have improved when denoising techniques have been applied as a preprocessing step.