Search

Remote sensing and digital agriculture have emerged as practical tools in modern farming, promising significant advancements in agricultural efficiency, sustainability, and productivity. These technologies could enable farmers and researchers to gather essential data, make informed decisions, and manage resources more effectively, ultimately contributing to the global challenge of feeding a growing population while minimizing environmental impact. Remote sensing data, especially high-resolution UAS-based imagery, could provide detailed and real-time information about crop health, soil moisture, nutrient levels, and pest infestations. This information play an important role in addressing key agricultural challenges including I) limited resources, II) newly emerging or common diseases, iii) climate change and extreme weather outbreaks, and IV) sustainability. The wealth of data enables farmers to implement precision agriculture techniques, tailoring interventions and resources to specific areas of the field. The timely detection and monitoring of crop diseases allows farmers to take proactive measures and prevent large-scale outbreaks. Early-stage detection of problems enable farmers to mitigate losses, improve crop quality, and reduce the need for excessive chemical treatments. Remote sensing data enables high-thruput phenotyping and selection of more resilient and climate-adept crop varieties. Finally, constant monitoring and precise resource management facilitate sustainable production. All advantages of remote sensing applications discussed above depends on the accuracy and reliability of the data collected and correct interpretation of the data.Despite the extensive deployment of remote sensing tools in agricultural industry for several decades, the anticipated outcomes have often fallen short of initial expectations. This shortcoming can be attributed to several interrelated factors including challenges related to data accuracy, resolution, data interpretation issues, and inconsistency across different spatial and temporal scales. Additionally, the dynamic nature of environmental processes and the inherent complexity of ecosystems makes it challenges to establish cause-and-effect relationships between remote sensing and ground truth data. The aforementioned challenges and limitations motivated our first study, which aims to comprehensively review the current platforms, applications, and methodologies employed in remote sensing (with a focus on nut crops) and identify the potential error sources that cause these shortcomings. In this study we realized that radiometric correction, which is an essential data preprocessing step, has not been taken seriously in most studies. Without correct radiometric calibration, the recorded values may not represent true surface reflectance, leading to inaccuracies in the quantitative analysis. Additionally, inaccurate calibration can hinder the ability to compare and integrate data from different sensors or acquired at different times, as the variations in sensor responses or environments may not be properly accounted for, hence preventing generalizability of the methods and results. Further investigation into the radiometric calibration steps and associated issues highlighted a significant environmental factor that can impact the collected data and lead to questionable interpretations. Specifically, the relationship between sun-plant-sensor geometry and the acquired remote sensing data stood up as a crucial discrepancy source. Conventionally, researchers collect reflectance data around “solar noon” in an attempt to minimize the irradiance changes. However, solar noon could be very different in terms of sun angle depending on the location and season. This formed our main study which focuses on understanding the impact of sun-view geometry on spectral reflectance variability of crops. In this study we showed that even a small 2° change in view angle of the remote sensing camera can lead to substantial differences in canopy reflectance, especially close to hotspot area. These variations, often assumed negligible, can exceed ± 50% of the nadir view due to directional solar radiation within a drone image. We introduced a model based on the Laplacian distribution function that can be used to compensated for these variations with up to %88 accuracy. With this model data become more comparable and hence the results more generalizable. Additionally, several other factors that are critical for radiometric calibration were investigated and recommendations for more reliable data collection or processing were suggested. Finally we developed an application, which is publicly accessible through Digital Ag Lab’s website (https://digitalag.ucdavis.edu/decision-support-tools/when2fly), recommends the best UAS flight time to collect more reliable agricultural remote sensing data based on the sites location, date, and camera’s field of view.

Mangrove forests are critical to mitigating climate change and provide many essential benefits to their ecosystems and local environments but are under threat due to deforestation. However, monitoring mangroves through remote sensing can help pinpoint and alleviate the causes of their deforestation. Machine learning can be used with remotely sensed low-resolution satellite or high-resolution aerial imagery to automatically create mangrove extent maps with higher accuracy and frequency than previously possible. This study explores and offers recommendations for two practical scenarios. In the first practical scenario, where only low-resolution hyperspectral satellite imagery is acquired, we implemented several classical machine learning models and applied these results to data acquired in the Clarendon parish of Jamaica. We found that utilizing extensive feature engineering and hyperspectral bands can result in strong performance for mangrove extent classification, with an accuracy of 93% for our extremely randomized trees model. In the second practical scenario, we explored when there is full coverage of both low-resolution satellite and high-resolution aerial imagery over a survey area. We created a hybrid model which fuses low-resolution pixels and high-resolution imagery, achieving an accuracy of 97% when applied to a dataset based in Baja California Sur, Mexico, offering another high-performance method to automatically create mangrove extent maps if both high- and low-resolution imagery is available. Overall, the methods tested over these two scenarios provide stakeholders flexibility in data and methods used to achieve accurate, automatic mangrove extent measurement, enabling more frequent mangrove monitoring and further enabling the protection of these important ecosystems.

The increasing trend of active wildfire events in California poses risks to both wildland ecosystems and human communities, causing substantial ecological damages and economic losses. However, the mechanism of wildfire damage within vegetated landscapes and human settlements has been understudied, especially for the effects of changing climates and the expanding wildland urban interfaces. Recent advances in remote sensing and machine learning enabled fine-scale observation of land changes from wildfire attacks and an improved analysis of environmental and anthropogenic factors controlling the wildfire damage. This dissertation has sought to bridge the knowledge gaps of wildfire risks across the land through remote sensing and machine learning approaches. Specifically, this research has aimed to 1) disentangle effects of environmental factors on burn severity under various extreme climatic conditions in California’s northern coastal mountains and predict near-real-time burn severity from remote sensing images and geospatial data; 2) develop an accurate and effective framework of biannual analysis of building footprints and wildland urban interface mapping from high resolution aerial photos; 3) perform the first comprehensive analysis of factors determining building damage from wildfire events at the daily scale across California. The results of this dissertation highlighted the importance of targeting areas with high burn severity or building damage risk for fire adaptation and mitigation strategies in a changing climate and expanding human settlement. These findings, when combined with building footprints and wildland urban interface products developed through a modeling framework, contribute to the guidance of land use planning to create fire- adaptive and resilient communities in California.

Advancements in machine learning hold unprecedented potential to help humans understand and shape our world, from deriving data-driven policies with global prediction systems to combating long-standing prejudice in social decision-making. However, in practice, machine learning is not achieving its potential. Algorithmic systems that are theoretically well motivated fail to live up to anticipated performance in the real world, and all too often, they exacerbate inequality rather than relieve it. In order to translate potential into real benefit, machine learning systems need to address the context of the domain they’re applied in to better interface with system intent and system impact.In this thesis, we present research on the context-aware design of machine learning systems that interface with individuals, our environment, and our societies. We emphasize the core themes of intent, impact, and context through three interweaving threads: (i) contrasting algorithmic impact with desired intent, (ii) contextualizing learning algorithms through structures in input data, and (iii) advancing machine learning with remotely sensed data as precise applications of intent- and context-aware design in practice. We conclude with overarching lessons learned that carry into constructive opportunities for future research.

There are many environments on Earth that are so remote that they are inhospitable to humans and conventional sensing equipment. Yet, these environments can hold information of ecological and cultural significance that cannot be gathered anywhere else. Current methods of gathering information in these environments give an important window, but utilizing modern sensors to capture 3D information can allow us to interpret existing data and understand the environments in new and unique ways. This thesis will demonstrate how 3D capture can improve data collection and interpretation in three separate remote environments. First, I will show how Synthetic Aperture Sonar on autonomous underwater vehicles paired with optimized feature detectors can improve target detection and seafloor recognition. Next, I will show how RGBD cameras, photogrammetry, and LIDAR can be used in isolated Guatemalan archaeological excavations to visualize and contextualize ancient sites in relation to each other and to our broader understanding of Mayan history. Finally, I will demonstrate the effectiveness and potential of RGBD cameras for fish stock assessment through detection and length and biomass measurement in open waters and in aquaculture.

Among all natural phenomena, precipitation is the main driver of the hydrological cycle and the challenging task of precipitation estimation is an essential element for hydrological and meteorological applications. Recent developments in satellite technologies resulting in higher temporal, spatial, and spectral resolutions, along with advancements in Machine Learning (ML) algorithms and computational power, open the great opportunity to develop analytical and data-driven tools to characterize such natural phenomena and their future behavior more efficiently and accurately. In this dissertation, state-of-the-art data-driven frameworks are proposed based on advanced deep learning algorithms and computer vision tools to extract the most useful features from single or multiple spectral bands of satellite information.Specifically, in the first part of the dissertation, a novel gradient-based cloud segmentation algorithm is proposed to effectively identify clouds and monitor their evolution towards more accurate quantitative precipitation estimation and forecast. This algorithm integrates morphological image gradient magnitudes to separate cloud systems and patches boundaries from single or multi-spectral imagery. This method improves rain detection and estimation skills with an accuracy rate of up to 98\% in identifying cloud regions compared to the existing cloud-patch-based segmentation technique implemented in the operational PERSIANN-CCS (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Network - Cloud Classification System) algorithm. Application of this method is extendable to hurricanes simulations and synthetic satellite imageries simulated by high-resolution weather prediction models. In the second part, an end-to-end deep learning precipitation estimation framework is developed from multiple sources of remotely sensed information to provide half-hourly, 4-km by 4-km precipitation estimates over the CONUS. In the first stage, a Rain/No Rain (R/NR) binary mask is generated by classification of the pixels and then a Fully Convolutional Network (FCN), U-net, is used as a regressor to predict precipitation estimates for rainy pixels. Due to the complex structure of precipitation, and inability of traditional objective functions to capture the true rainfall distribution, a novel distribution matching approach is designed and implemented. The network is trained using both the conditional Generative Adversarial Network (cGAN), and the Mean Squared Error (MSE) loss terms to match the distribution of the generated results and observed data, and to relax the strict assumptions of the traditional and conventional loss function in DNNs. The newly developed precipitation estimation algorithm is introduced as an augmentation for PERSIANN-CCS (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Network - Cloud Classification System) algorithm for estimating global precipitation and is termed as PERSIANN-cGAN. Statistics and visualizations of the metrics for PERSIANN-cGAN represent an improvement of the precipitation retrieval accuracy compared to the operational PERSIANN-CCS product, and a baseline model trained using the conventional MSE loss term.

With the development of launch vehicles, the number of satellite-based synthetic aperture radar (SAR) has been increasing. Satellite-based SAR can acquire high-resolution images at day and night regardless of the weather. Since SAR resolution is determined by a chirp bandwidth, corresponding a generation technology of wideband chirp is essentially required for high-resolution SAR images under consideration of relatively small and light weight SAR system. Therefore, an ideal chirp waveform generation, which is challenging task in actual hardware, is proposed by modifying a chirp equation for phase calibration, and is verified based on Sentinel-1 satellite SAR parameters. The proposed method is applied to our SAR hardware and we confirm the calibrated waveform. Evaluation results using SAR observation simulation show that the impulse response with the proposed method has better performances.

Remote sensing image data clustering is a tough task, which involves classifying the image without any prior information. Remote sensing image clustering, in essence, belongs to a complex optimization problem, due to the high dimensionality and complexity of remote sensing imagery. Therefore, it can be easily affected by the initial values and trapped in locally optimal solutions. Meanwhile, remote sensing images contain complex and diverse spatial-spectral information, which makes them difficult to model with only a single objective function. Although evolutionary multiobjective optimization methods have been presented for the clustering task, the tradeoff between the global and local search abilities is not well adjusted in the evolutionary process. In this article, in order to address these problems, a multiobjective sine cosine algorithm for remote sensing image data spatial-spectral clustering (MOSCA_SSC) is proposed. In the proposed method, the clustering task is converted into a multiobjective optimization problem, and the Xie-Beni (XB) index and Jeffries-Matusita (Jm) distance combined with the spatial information term (SI_Jm measure) are utilized as the objective functions. In addition, for the first time, the sine cosine algorithm (SCA), which can effectively adjust the local and global search capabilities, is introduced into the framework of multiobjective clustering for continuous optimization. Furthermore, the destination solution in the SCA is automatically selected and updated from the current Pareto front through employing the knee-point-based selection approach. The benefits of the proposed method were demonstrated by clustering experiments with ten UCI datasets and four real remote sensing image datasets.

Scene classification of high spatial resolution (HSR) images can provide data support for many practical applications, such as land planning and utilization, and it has been a crucial research topic in the remote sensing (RS) community. Recently, deep learning methods driven by massive data show the impressive ability of feature learning in the field of HSR scene classification, especially convolutional neural networks (CNNs). Although traditional CNNs achieve good classification results, it is difficult for them to effectively capture potential context relationships. The graphs have powerful capacity to represent the relevance of data, and graph-based deep learning methods can spontaneously learn intrinsic attributes contained in RS images. Inspired by the abovementioned facts, we develop a deep feature aggregation framework driven by graph convolutional network (DFAGCN) for the HSR scene classification. First, the off-the-shelf CNN pretrained on ImageNet is employed to obtain multilayer features. Second, a graph convolutional network-based model is introduced to effectively reveal patch-to-patch correlations of convolutional feature maps, and more refined features can be harvested. Finally, a weighted concatenation method is adopted to integrate multiple features (i.e., multilayer convolutional features and fully connected features) by introducing three weighting coefficients, and then a linear classifier is employed to predict semantic classes of query images. Experimental results performed on the UCM, AID, RSSCN7, and NWPU-RESISC45 data sets demonstrate that the proposed DFAGCN framework obtains more competitive performance than some state-of-the-art methods of scene classification in terms of OAs.

The normal growth of fishes is closely relevant to the density of mariculture. It is of great significance to accurately calculate the breeding area of specific sea area from satellite remote sensing images. However, there are no reports about cage segmentation and density detection based on remote sensing images so far. And the accurate segmentation of cages faces challenges from very large high-resolution images. Firstly, a new public mariculture cage data set is built. Secondly, the training set is augmented via sample variations to improve the robustness of the model. Then, for cage segmentation and density statistics, a new methodology based on Mask R-CNN is proposed. Using dividing and stitching technologies, the entire remote sensing test images of the cage can be accurately segmented. Finally, using the trained model, the object detection features and segmentation characteristics can be obtained at the same time. Considering only the area within the target detection frame, the proposed method can count the pixels in the segmented area, which can obtain accurate area and density while reducing time-consuming. Experimental results demonstrate that, compared with traditional contour extraction method and U-Net based scheme, the proposed scheme can significantly improve segmentation precision and model’s robustness. The relative error of the actual area is only 1.3%.

Currently, hospitals and health care sectors employ low-cost IoT (Internet of Thing) based remote health monitoring systems and labs in order to collect a subjects or patients real-time data. Such a process can be helpful for the early detection of a healthy newborn life and even vital in the case of saving these lives. In this paper, a preliminary implementation of a system monitoring the fetus heart rate has been designed and implemented as a mobile wearable measuring system with remote sensing. The proposed implementation turns out to be an efficient combination of simplicity and cost effectiveness and is accompanied with preliminary accurate measurements of fetus heart rate. The developed board uses an ESP-8266 transceiver module and is programmed to transmit the data efficiently to a remote server station using a IEEE 802.11 b/g/n - based wireless network. The patients data can then be monitored using an Android-based or satellite phone, or even any well-known internet browser connected to the specific network certifying in many occasions the health safety distance required due to various situations including that of COVID-19. IEEE

In this paper, we present SoDa, an irradiance-based synthetic Solar Data generation tool to generate realistic sub-minute solar photovoltaic (PV) output power time series, that emulate the weather pattern for a certain geographical location. Our tool relies on the National Solar Radiation Database (NSRDB) to obtain irradiance and weather data patterns for the site. Irradiance is mapped onto a PV model estimate of a solar plant’s 30-min power output, based on the configuration of the panel. The working hypothesis to generate high-resolution (e.g. 1 second) solar data is that the conditional distribution of the time series of solar power output given the cloud density is the same for different locations. We therefore propose a stochastic model with a switching behavior due to different weather regimes as provided by the cloud type label in the NSRDB, and train our stochastic model parameters for the cloudy states on the high-resolution solar power measurements from a Phasor Measurement Unit (PMU). In the paper we introduce the stochastic model, and the methodology used for the training of its parameters. The numerical results show that our tool creates synthetic solar time series at high resolutions that are statistically representative of the measured solar power and illustrate how to make use of the tool to create synthetic data for arbitrary sites in the footprint covered by the NSRDB.

This paper addresses the issue of the satellite image processing using GRASS GIS in the mangrove forests of the Niger River Delta, southern Nigeria. The estuary of the Niger River Delta in the Gulf of Guinea is an essential hotspot of biodiversity on the western coast of Africa. At the same time, climate issues and anthropogenic factors affect vulnerable coastal ecosystems and result in the rapid decline of mangrove habitats. This motivates monitoring of the vegetation patterns using advanced cartographic methods and data analysis. As a response to this need, this study aimed to calculate and map several vegetation indices (VI) using scripts as advanced programming methods integrated in geospatial studies. The data include four Landsat 8-9 OLI/TIRS images covering the western segment of the Niger River Delta in the Bight of Benin for 2013, 2015, 2021, and 2022. The techniques included the ’i.vi’, ’i.landsat.toar’ and other modules of the GRASS GIS. Based on the GRASS GIS ’i.vi’ module, ten VI were computed and mapped for the western segment of the Niger River Delta estuary: Atmospherically Resistant Vegetation Index (ARVI), Green Atmospherically Resistant Vegetation Index (GARI), Green Vegetation Index (GVI), Difference Vegetation Index (DVI), Perpendicular Vegetation Index (PVI), Global Environmental Monitoring Index (GEMI), Normalized Difference Water Index (NDWI), Second Modified Soil Adjusted Vegetation Index (MSAVI2), Infrared Percentage Vegetation Index (IPVI), and Enhanced Vegetation Index (EVI). The results showed variations in the vegetation patterns in mangrove habitats situated in the Niger River Delta over the last decade as well as the increase in urban areas (Onitsha, Sapele, Warri and Benin City) and settlements in the Delta State due to urbanization. The advanced techniques of the GRASS GIS of satellite image processing and analysis enabled us to identify and visualize changes in vegetation patterns. The technical excellence of the GRASS GIS in image processing and analysis was demonstrated in the scripts used in this study.

This research presents a 3D geographic information systems (GIS) modeling approach at a fine spatiotemporal resolution to assess solar potential for the development of smart net-zero energy communities. It is important to be able to accurately identify the key areas on the facades and rooftops of buildings that receive maximum solar radiation, in order to prevent losses in solar gain due to obstructions from surrounding buildings and topographic features. A model was created in ArcGIS, in order to efficiently compute and iterate the hourly solar modeling and mapping process over a simulated year. The methodology was tested on a case study area located in southern Ontario, where two different 3D models of the site plan were analyzed. The accuracy of the work depends on the resolution and sky size of the input model. Future work is needed in order to create an efficient iterative function to speed the extraction process of the pixelated solar radiation data.

The study involves an object-based segmentation method to extract feature changes in tropical rainforest cover using Landsat image and airborne LiDAR (ALS). Disturbance event that are represents the changes are examined by the classification of multisensor data; that is a highly accurate ALS with different resolutions of multispectral Landsat image. Disturbance Index (DI) derived from Tasseled Cap Transformation, Normalized Difference Vegetation Index (NDVI), and the ALS height are the variables for object-based segmentation process. The classification is categorized into two classes; disturbed and non-disturbed forest cover using Nearest Neighbor (NN), Random Forest (RF) and Support Vector Machine (SVM). The overall accuracy ranging from 88% to 96% and kappa ranging from 0.79 to 0.91. Mcnemar’s test p-value (

This paper presents a survey of image processing techniques proposed in the literature for extracting key cereal crop growth metrics from high spatial resolution, typically proximal images. The descriptive crop growth metrics considered are: crop canopy cover, above ground biomass, leaf area index (including green area index), chlorophyll content, and growth stage. The paper includes an overview of relevant fundamental image processing techniques including camera types, colour spaces, colour indexes, and image segmentation. The descriptive crop growth metrics are defined. Reference methods for ground-truth measurement are described. Image processing methods for metric estimation are described in detail. The performance of the methods is reviewed and compared. The survey reveals limitations in image processing techniques for cereal crop monitoring such as lack of robustness to lighting conditions, camera position, and self-obstruction. Directions for future research to improve performance are identified.

Due to the attractive potential in avoiding the elaborate definition of anchor attributes, anchor-free-based deep learning approaches are promising for object detection in remote sensing imagery. CornerNet is one of the most representative methods in anchor-free-based deep learning approaches. However, it can be observed distinctly from the visual inspection that the CornerNet is limited in grouping keypoints, which significantly impacts the detection performance. To address the above problem, a novel and effective approach, called GroupNet, is presented in this paper, which adaptively groups corner specific to the objects based on corner embedding vector and corner grouping network. Compared with the CornerNet, the proposed approach is more effective in learning the semantic relationship between corners and improving remarkably the detection performance. On NWPU dataset, experiments demonstrate that our GroupNet not only outperforms the CornerNet with an AP of 12.8%, but also achieves comparable performance to considerable approaches with 83.4% AP.

An increase in radiotherapy-induced secondary malignancies has led to recent developments in analytical modelling of out-of-field dose. These models must be validated against measurements, but currently available datasets are outdated or limited in scope. This study aimed to address these shortcomings by producing a large dataset of out-of-field dose profiles measured with modern equipment. A novel method was developed with the intention of allowing physicists in all clinics to perform these measurements themselves using commonly available dosimetry equipment. A standard 3D scanning water tank was used to collect 36 extended profiles. Each profile was measured in two sections, with the inner section measured with the beam directly incident on the tank, and the outer section with the beam incident on a water-equivalent phantom abutted next to the tank. The two sections were then stitched using a novel feature-matching approach. The profiles were compared against linac commissioning data and manually inspected for discontinuities in the overlap region. The dataset is presented as a publicly accessible comma separated variable file containing off-axis ratios at a range of off-axis distances. This dataset may be applied to the development and validation of analytical models of out-of-field dose. Additionally, it may be used to inform dose estimates to radiosensitive implants and anatomy. Physicists are encouraged to perform these out-of-field measurements in their own clinics and share their results with the community., 22 pages, 9 figures; Updated structure for submission to new journal

As one of the most relevant data preprocessing techniques, discretization has played an important role in data mining, which is widely applied in industrial control. It can transform continuous features to discrete ones, thus improving the efficiency of data processing and adapting to learning algorithms that require discrete data as inputs. However, traditional discretization methods have shortcomings, such as highly complex programs, excessive numbers of intervals obtained, and significant loss of necessary information in the preprocessing of high-resolution remote sensing big data. Moreover, the large number of mixed pixels in the image is a primary reason for the uncertainty of remote sensing information systems, and current discretization methods are based on the assumption that one pixel only corresponds to the spectral information of a single object, without considering the influence of the uncertainty caused by a mixed spectrum, which causes the classification accuracy to drop after discretization. We propose a discretization method for high-resolution remote sensing big data. We determine the membership degree of each pixel in training samples through linear decomposition, and establish the individual fitness function based on a fuzzy rough model. An adaptive genetic algorithm selects discrete breakpoints, and a MapReduce framework calculates the individual fitness of the population in parallel, to obtain the optimal discretization scheme in the minimum time. Our method is compared to the best state-of-the-art discretization algorithms on the authentic remote sensing datasets. Experiments verified the effectiveness of the proposed method, which provides strong support for the subsequent processing of images. © 2020 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

Deep learning models have become the mainstream algorithm for processing computer vision tasks. In object detection tasks, the detection box is usually set as a rectangular box aligned with the coordinate axis, so as to achieve the complete package of the object. However, when facing some objects with large aspect ratio and angle, the bounding box has to become large, which makes the bounding box contain a large amount of useless background information. In this study, a different approach is taken, using a method based on YOLOv5, the angle information dimension is increased in head part and angle regression added at the same time of the border regression, combining ciou and smoothl1 to calculate the bounding box loss, so that the resulting border box fits the actual object more closely. At the same time, the original dataset tags are also preprocessed to calculate the angle information of interest. The purpose of these improvements is to realize object detection with angles in remote-sensing images, especially for objects with large aspect ratios, such as ships, airplanes, and automobiles. Compared with the traditional object detection model based on deep learning, experimental results show that the proposed method has a unique effect in detecting rotating objects.

We propose a data-driven method for simulating lidar sensors. The method reads computer-generated data, and (i) extracts geometrically simulated lidar point clouds and (ii) predicts the strength of the lidar response - lidar intensities. Qualitative evaluation of the proposed pipeline demonstrates the ability to predict systematic failures such as no/low responses on polished parts of car bodyworks and windows, or strong responses on reflective surfaces such as traffic signs and license/registration plates. We also experimentally show that enhancing the training set by such simulated data improves the segmentation accuracy on the real dataset with limited access to real data. Implementation of the resulting lidar simulator for the GTA V game, as well as the accompanying large dataset, is made publicly available.

Labeling of the connected components is the key operation of the target recognition and segmentation in remote sensing images. The conventional connected-component labeling (CCL) algorithms for ordinary optical images are considered time-consuming in processing the remote sensing images because of the larger size. A dynamic run-length based CCL algorithm (DyRLC) is proposed in this paper for the large size, big granularity sparse remote sensing image, such as space debris images and ship images. In addition, the equivalence matrix method is proposed to help design the pre-processing method to accelerate the equivalence labels resolving. The result shows our algorithm outperforms 22.86% on execution time than the other algorithms in space debris image dataset. The proposed algorithm also can be implemented on the field programming logical array (FPGA) to enable the realization of the real-time processing on-board.

Sea fog detection with remote sensing images is a challenging task. Driven by the different image characteristics between fog and other types of clouds, such as textures and colors, it can be achieved by using image processing methods. Currently, most of the available methods are data-driven and relying on manual annotations. However, because few meteorological observations and buoys over the sea can be realized, obtaining visibility information to help the annotations is difficult. Considering the feasibility of obtaining abundant visible information over the land and the similarity between land fog and sea fog, we propose an unsupervised domain adaptation method to bridge the abundant labeled land fog data and the unlabeled sea fog data to realize the sea fog detection. We used a seeded region growing module to obtain pixel-level masks from rough-labels generated by the unsupervised domain adaptation model. Experimental results demonstrate that our proposed method achieves an accuracy of sea fog recognition up to 99.17%, which is nearly 3% higher than those vanilla methods.

Low-altitude communication has attracted lots of attention in recent years, and it is very important to obtain an accurate clutter loss model for evaluating and designing such a communication system. For this purpose, we have conducted a channel measurement campaign at 5.8 GHz for suburabn low-altitude scenario. Based on the measured pathloss, we extract and analyze the clutter loss for different link distances and elevation angles between the transceiver. The measured clutter loss is validated by the ITU clutter loss model with respect to the link distance. Moreover, we compare the measured clutter loss with the the ITU clutter loss model with respect to the elevation angle, and find the obvious difference between them. Therefore, in this letter, we propose a modified clutter loss model and a pathloss model with respect to the elevation angle factor for the suburban low-altitude scenario.

Remote sensing image scene classification has attracted great attention because of its wide applications. Although convolutional neural network (CNN)-based methods for scene classification have achieved excellent results, the large-scale variation of the features and objects in remote sensing images limits the further improvement of the classification performance. To address this issue, we present multiscale representation for scene classification, which is realized by a global-local two-stream architecture. This architecture has two branches of the global stream and local stream, which can individually extract the global features and local features from the whole image and the most important area. In order to locate the most important area in the whole image using only image-level labels, a weakly supervised key area detection strategy of structured key area localization (SKAL) is specially designed to connect the above two streams. To verify the effectiveness of the proposed SKAL-based two-stream architecture, we conduct comparative experiments based on three widely used CNN models, including AlexNet, GoogleNet, and ResNet18, on four public remote sensing image scene classification data sets, and achieve the state-of-the-art results on all the four data sets. Our codes are provided in https://github.com/hw2hwei/SKAL.