Search

Traditional target detection algorithms have difficulty to adapt complex environmental changes and have limited applicable scenarios. However, the deep-learning-based target detection model can automatically learn with strong generalization capability. In this article, we choose a single-stage deep-learning-based target detection model for research based on the model’s real-time processing requirements and to improve the accuracy and the robustness of target detection in remote sensing images. In addition, we improve the YOLOv4 network and present a new approach. First, we propose a classification setting of the nonmaximum suppression threshold to increase the accuracy without affecting the speed. Second, we study the anchor frame allocation problem in YOLOv4 and propose two allocation schemes. The proposed anchor frame scheme also improves the detection performance, and experimental results on the DOTA dataset validate their effectiveness.

A geostationary microwave sounder, capable of providing continuous monitoring of temperature, water vapor, clouds, precipitation, and wind in the presence of clouds and precipitation is now feasible. A design called the Geostationary Synthetic Thinned Aperture Radiometer (GeoSTAR) has been developed at the Jet Propulsion Laboratory, and the required new technology has been developed and is sufficiently mature that a space mission can be initiated. GeoSTAR can be thought of as “AMSU in GEO,” i.e., it has capabilities in geostationary earth orbit (GEO) similar to those of microwave sounders currently operating in low earth orbit. Having such a capability in GEO will add tremendously to our ability to observe dynamic atmospheric phenomena, such as hurricanes and severe storms, monsoonal moisture flow, and atmospheric rivers. GeoSTAR will make measurements every 15 min or less instead of every 12 h and cover a large portion of the Earth continuously instead of with snapshots in a narrow swath. By tracking water vapor patterns, it is also possible to derive atmospheric wind speed and direction at altitudes from the surface to 10–15 km. All of this can be done regardless of cloud cover and weather conditions. During the latter half of 2020, a detailed study of GeoSTAR and its projected performance was undertaken as one of several such studies commissioned by the National Oceanic and Atmospheric Administration (NOAA) for the purpose of configuring NOAA's next generation of earth environmental satellite systems. We present a summary of our findings, including instrument characteristics, measurement accuracy and precision, and expected impact on weather prediction and applications.

Ship detection and classification in synthetic aperture radar (SAR) images play a vital role for wide applications. Due to the unique SAR imaging mechanism, ship detection and classification tasks have faced numerous challenges, such as land interference, image defocus, and noise. Many detectors and classifiers have been presented to handle these problems. However, the general deep learning-based detectors and classifiers lack the combination of SAR characteristics, which leads to poor performance. Compared with optical images, SAR images lack the texture information of ships, which brings great difficulties to the recognition task. To address the above issues, a novel deep learning-based ship detection and classification network combined with scattering characteristics is proposed in this article. First, to accurately locate ships in large-scale SAR images, this article designs a strong scattering point aware network (SPAN) by capturing the strong scattering points that existed in the ship area. SPAN recognizes the ship category according to their distribution characteristics. Second, to compensate for the feature loss caused by the down-sampling operation, this article designs a more suitable resolution recovery module to replace the bilinear interpolation method. Third, a region of interest automatic generation module is proposed to fully utilize the axis-align feature of oriented proposal boxes and the sufficient information of horizontal proposal boxes. Furthermore, the classification encoder module extracts the distribution feature of scattering points to classify SAR ships. Finally, the comprehensive experiments in the large-scale dataset for ship detection and classification in SAR images (LDSD) demonstrate the superior performance of the proposed method.

Compact polarimetry (CP) has attracted much attention in recent years due to its hybrid dual-polarization imaging mode. CP synthetic aperture radar has a larger swath and can provide more polarimetric information compared with the traditional dual-polarization imaging mode (HH/HV or VH/VV). Pseudo quad-polarimetric (quad-pol) data reconstruction is an important technology in the application of CP data. The goal of pseudo quad-pol data reconstruction from CP data is to change the form of CP data to the form of quad-pol data without increasing any new information. In this article, a new pseudo quad-pol data reconstruction method from the CP data is proposed. This method combines a complex-valued dual-branch convolutional neural network (CV-DBCNN) to achieve the reconstruction of the pseudo quad-pol data. It utilizes complex-valued convolutional layers and a complex-valued activation function to fully extract the polarimetric information embedded in the complex-valued CP data. For the CV-DBCNN, the branch with 1×1 kernel size is used to nonlinearly and self-adaptively combine the channel of input data, and the branch with 3×3 kernel size is used to extract the discriminative regional polarimetric features. Furthermore, polarimetric decomposition is utilized to evaluate the scattering mechanisms of the pseudo quad-pol data. Three state-of-the-art methods are utilized for comparison. In comparison with other methods, our proposed reconstruction method based on the CV-DBCNN shows its superiority in terms of the pseudo quad-pol data reconstruction and scattering mechanism preservation.

Segmentation of marine oil spills in synthetic aperture radar (SAR) images is a challenging task because of the complexity and irregularities in SAR images. In this work, we aim to develop an effective segmentation method which addresses marine oil spill identification in SAR images by investigating the distribution representation of SAR images. To seek effective oil spill segmentation, we revisit the SAR imaging mechanism in order to attain the probability distribution representation of oil spill SAR images, in which the characteristics of SAR images are properly modelled. We then exploit the distribution representation to formulate the segmentation energy functional, by which oil spill characteristics are incorporated to guide oil spill segmentation. Moreover, the oil spill segmentation model contains the oil spill contour regularization term and the updated level set regularization term which enhance the representational power of the segmentation energy functional. Benefiting from the synchronization of SAR image representation and oil spill segmentation, our proposed method establishes an effective oil spill segmentation framework. Experimental evaluations demonstrate the effectiveness of our proposed segmentation framework for different types of marine oil spill SAR image segmentation.

As discussed in the previous part of this review article, remote sensing (RS) creates unprecedented opportunities by providing a variety of systems with different characteristics to study and monitor oceans. Part 1 of this review article was dedicated to reviewing passive RS systems and their main applications in the ocean. Here, in part 2, seven active RS systems, including scatterometers, altimeters, gravimeters, synthetic aperture radar, light detection and ranging, sound navigation and ranging, high-frequency radars are comprehensively reviewed. For consistency, this part is structured similarly to part 1. The aforementioned systems, along with their characteristics and primary applications, are introduced in separate sections. This review article provides useful information to all students and researchers who are interested in the oceanographic applications of active RS systems.

The upper Luanhe River Basin is a significant ecological barrier guarding the Beijing–Tianjin–Hebei region in China. Quantitative measures of vegetation productivity can be used to assess ecosystem carbon sequestration capacity and monitor regional ecological environmental health. Although several vegetation productivity products have been generated, poor spatiotemporal resolution limits their application in ecosystem service assessment. In this article, vegetation net primary productivity (NPP) from 2000 to 2017 with a resolution of 30 m in the upper Luanhe River Basin was generated based on a data fusion model and the multisource data synergized quantitative (MuSyQ) NPP model. Then, the variation trend of NPP and its climate controls were analyzed. Compared with forest NPP observation data, we derived an R2 of 0.68 and the root-mean-square error of 81.70 gC.m−2.yr−1. Annual NPP had a fluctuating increasing trend from 2001 to 2017, with values ranging between 3.43 and 5.00 TgC.yr−1, with an annual increase trend of 0.04 TgC.yr−1. Precipitation was significantly correlated with NPP in the upper part of the Luanhe River basin, which is an important reason for the interannual variation of NPP. Grassland had a stronger correlation to precipitation than forest because it is more sensitive to precipitation. The area where the temperature is significantly correlated with annual NPP only accounts for 2% of the study area, indicating that temperature has a weak influence on NPP. Furthermore, human activities, such as forest management, fertilization, and irrigation, can change the trend of annual NPP.

Snow liquid water produced by melting can affect the surface mass and energy balance in the Antarctic Ice Sheet (AIS). It is essential for monitoring the occurrence of snow liquid water (OLW) within the snowpack. Spaceborne microwave sensors (i.e., scatterometer and radiometer) and climate models are primary tools for examining the OLW in the AIS. However, implementing the complementary nature of measurements and comparing their observations quantitatively are challenging owing to sparse snow wetness validation data. In this article, we use categorical triple collocation to rank the relative performance of OLW products. According to the first rankings, we construct an ensemble OLW product that is more consistent with the results detected by in situ station air temperature data than single data source products. Furthermore, we quantitatively estimate the proportion correct of wet snow (sensitivity) and the proportion correct of dry snow (specificity) of the measurements on the Antarctic Peninsula (AP). The scatterometer demonstrates high balanced accuracy (a binary-variable performance metric) on the AP (up to 0.899); however, this result does not signify that the scatterometer is the optimal observation in any period. The radiometer and climate model show high sensitivities in different melt stages but underestimate wet snow extent during certain periods. The quantitative estimation provides a new perspective for comparing various observations and detection methods.

Accurate large-scale building detection is significant in monitoring urban development, map updating, change detection, and digital city establishment. However, due to the complicated details of background objects in high spatial resolution remotely sensed images, the models proposed in building detection are still not performing satisfactorily. Particularly, such issue lies in the small buildings, which are easily to be omitted, and the pixels in the bounding area of each building instance can be especially confusing with the background objects. Aiming to deal with such problem, we propose Res2-Unet to employ multi-scale learning at a granular level, rather than the commonly used layer-wise feature learning, to enlarge the scale of receptive fields of each bottleneck layer. It replaces the widely used 3 $ \times $ 3 convolution on n-channel feature maps with a set of smaller groups, which are organized in a hierarchical structure to enlarge the scale-variability. The general framework is an end-to-end learning network, taking a typical semantic segmentation network structure with encoders to encode the input image into feature maps and decoders to decode the feature maps into binary segmented result image. Moreover, to enhance the building boundary generation ability of our model, a boundary loss function is proposed to improve the detection performance. The proposed framework is evaluated on three public datasets, Massachusetts building dataset, WHU East Asia Satellite dataset and WHU Aerial building dataset. It is compared with the published performances and has achieved the state-of-the-art accuracies. That verifies the robustness of the proposed framework.

Recent decades have witnessed a tremendous growth in the number of Earth observation satellites (EOSs), which presents a huge challenge for mission planning. For the EOSs with optical sensors particularly, the observation mission is significantly influenced by the uncertainty of cloud coverage, which has been identified as the most dominant factor for the invalidation of remote sensing images. To overcome this uncertainty, uncertainty programming methods, namely, chance constraint programming (CCP), stochastic expectation model, and robust optimization, are put forth. Despite their success, these approaches are limited in that they simplified the complex cloud coverage uncertainty, which may be different from the true cloud conditions, and they did not take the true cloud information into consideration. Motivated by these recent trends toward Big Data of satellite cloud images and machine learning for spatiotemporal prediction, this article explores a dynamic replanning scheme for multiple EOSs based on cloud forecasting. Specifically, we propose a new approach mainly in the following three steps: first, proactive scheduling based on a CCP is implemented and uploaded via ground control; second, cloud forecasting can be continuously conducted relying on the predictive recurrent neural network and the latest satellite cloud image; and third, mission replanning can be conducted according to the initial schedule and relatively accurate cloud information. Simulation results show that the cloud forecasting method is effective, and the replanning approach presents highly efficient and accurate scheduling results.

Remote sensing (RS) scene classification plays an important role in a wide range of RS applications. Recently, convolutional neural networks (CNNs) have been applied to the field of scene classification in RS images and achieved impressive performance. However, to classify RS scenes, most of the existing CNN methods either utilize the high-level features from the last convolutional layer of CNNs, missing much important information existing at the other levels, or directly fuse the features at different levels, bringing redundant and/or mutually exclusive information. Inspired by the attention mechanism of the human visual system, in this article, we explore a novel relation-attention model and design an end-to-end relation-attention network (RANet) to learn powerful feature representations of multiple levels to further improve the classification performance. First, we propose to extract convolutional features at different levels by pretrained CNNs. Second, a multiscale feature computation module is constructed to connect features at different levels and generate multiscale semantic features. Third, a novel relation-attention model is designed to focus on the critical features whilst avoiding the use of redundant and even distractive ones by exploiting the scale contextual information. Finally, the resulting relation-attention features are concatenated and fed into a softmax layer for the final classification. Experiments on four well-known RS scene classification datasets (UC-Merced, WHU-RS19, AID, and OPTIMAL-31) show that our method outperforms some state-of-the-art algorithms.

Monitoring glacier dynamics is an effective approach for quantifying the response of the glaciers in cold arid mountainous areas to climate change. However, the quantification of short-term flow dynamics of the mountain glaciers in cold-dry climates has rarely been reported. This article investigated the intra-annual flow dynamics in terms of velocity and ice thickness changes on the Kunlun glacier, a mountain glacier in the cold-dry west Kunlun mountains, using spaceborne synthetic aperture radar (SAR) imagery. We applied the improved pixel-offset-tracking small-baseline-subset method (PO-SBAS) on ascending and descending Sentinel-1A SAR images acquired in 2017 and 2018 to estimate the three-dimensional (i.e., north–south, west–east, and vertical) velocity time series of the glacier. The vertical velocities were further decomposed into the surface-parallel-flow (SPF) and the nonsurface-parallel-flow (nSPF) components, which link glacier motion along glacier surface slope and internal ice deformation, respectively, to glacier thickness changes. Our findings show that the eastern branch of the glacier moved faster than the western branch. We inferred that a loss of ice thickness due to a previous surge on the western branch should be responsible for its slower flow. The nSPF rates are higher than the SPF rates in both branches, indicating that internal ice deformation primarily controls the changes in ice thickness. We also observed an apparent summer acceleration in the nSPF rates, which is likely caused by changes in subglacial hydrological conditions. This article highlights the potential uses of the improved PO-SBAS method of quantifying the flow dynamics of the glaciers in cold-dry mountain regions.

Synthetic aperture radar automatic target recognition (SAR ATR) uses computer processing capabilities to infer the classes of the targets without human intervention. For SAR ATR, deep learning gradually emerges as a powerful tool and achieves promising performance. However, it faces serious challenges of how to deal with incremental recognition scenarios. The existing deep learning-based SAR ATR methods usually predefine the total number of recognition classes. In realistic applications, the new tasks/classes will be added continuously. If all old data are stored and mixed with newly added data to update the model, the storage pressure and time consumption make the application infeasible. In this article, the high plastic error correction incremental learning (HPecIL) is proposed to address the model degradation and plasticity decline in the incremental scenario. Multiple optimal models trained on old tasks are used to correct accumulative errors and alleviate model degradation. Moreover, the sharp data distribution shift due to newly added data can also result in the model underperforming. A class-balanced training batch is constructed to deal with the issue of unbalanced data distribution. To make a tradeoff between model stability and model plasticity, low-effect nodes in the model are removed to boost the efficiency of model update. The proposed HPecIL outperforms the other state-of-the-art methods in incremental recognition scenarios. The experimental results demonstrate the effectiveness of the proposed method.

This study assimilated Sentinel-1 C-band backscatter observations over snow-covered terrain into the Noah-Multiparameterization land surface model using support vector machine (SVM) regression and an ensemble Kalman filter to improve the modeled terrestrial snow mass estimates. The data assimilation (DA) experiment was conducted across Western Colorado from September 2016 to August 2017. As part of the DA experiments, the impact of a rule-based update was evaluated by comparing snow water equivalent (SWE) estimates via DA (with [${\rm {DA}}_{\rm {v1}}$] and without [${\rm {DA}}_{\rm {v2}}$] the rule-based update) against SNOTEL SWE measurements. Results confirmed that rule-based update helped minimize SVM controllability issues, and in turn, improved the accuracy of SWE estimates relative to both open loop (OL) and ${\rm {DA}}_{\rm {v2}}$. Comparison of SWE estimates from Sentinel-1 ${\rm {DA}}_{\rm {v1}}$ against SNOTEL SWE revealed that 75% of stations showed improvements in bias and correlation coefficient relative to the OL. Assimilated SWE estimates also showed statistical improvements during both the snow accumulation and snow ablation periods. However, unbiased root mean square error showed a slight increase during the snow ablation period due to the large variability in the electromagnetic response of C-band backscatter over deep and/or wet snow. Improvement of the SWE estimates also resulted in improving river discharge estimates compared to in situ measurements. River discharge using Sentinel-1 ${\rm {DA}}_{\rm {v1}}$ improved the Nash–Sutcliffe efficiency at all available stations. These results suggest that physically constrained SVM can serve as an efficient observation operator for snow mass DA through explicit consideration of the first-order C-band scattering mechanisms over different terrestrial snow conditions.

Extracting objects of interest from remote sensing imagery is an essential part in various practical applications. The objects that people pay attention to in the remote sensing scene mainly include buildings, roads, vehicles, etc. In this article, extracting the aforementioned objects are collectively referred to as the target extraction task. Arising from object scale variation, appearance similarity between adjacent patches, diversity of imaging orientation, and complexity of background, it is difficult to extract complete objects from cluttered backgrounds. Deep neural network has made great achievement in dense prediction for target extraction. However, most of the previous works are still faced with a formidable challenge in discriminative context feature representation to extract targets of various categories and correctly classify pixels around the boundary. In this article, we propose a target extraction neural network, named discriminative context-aware network, to focus on discriminative high-level context features and preserve spatial information. First, a discriminative context-aware feature module is designed to generate the feature maps in the top layer, which not only captures the rich image context information but also aggregates the contrasted local information at multiple scales. Second, a refine decoder module is adopted to preserve spatial information from low-level layers and enhance the feature representation, leading to precise segmentation results. We conducted extensive experiments on building and road extraction benchmarks, including WHU building dataset and Massachusetts road dataset, together with a self-constructed dataset for vehicle extraction in SAR images. Our method achieves state-of-the-art results with fewer parameters and faster inference.

This study presents the results of a field experiment conducted for assessing the crop health status of several barley and oat crop fields in Prince Edward Island, Canada. The crop fields were mapped with an unmanned aircraft system (UAS), and the crop health status was assessed through the green area index (GAI) and vegetation indices (VIs). GAI maps were produced from the UAS imagery and VIs used machine learning pipelines with several regression algorithms (multiple linear models, support vector machines, random forests, and artificial neural networks) along with a feature selection strategy. The random forests algorithm was shown to be the best algorithm for GAI prediction with an average relative root mean square error of 10.86% and a mean absolute error of 0.67. The resulting GAI maps and the regression feature space were classified with random forests to discriminate between vigorous and stressed crop areas. We achieved a mean overall accuracy of 94%. The limits of the study are also presented.

Video synthetic aperture radar (SAR) is attracting more and more attention because of its continuous imaging capability for ground scene of interest under any weather conditions and any time of the day. To reduce the sampling amount of video SAR, image processing can be formulated into a low-tubal-rank tensor recovery problem. In this article, we proposed an orthogonal rank-1 tensor pursuit (ORTP) algorithm to solve the low-tubal-rank tensor recovery problem in video SAR imaging. The proposed ORTP algorithm is an extension of the orthogonal rank-1 matrix pursuit algorithm in the matrix sensing problem from the matrix case to the tensor case under a tubal-rank model. It is capable of reconstructing the target tensor efficiently without requiring any prior information about the prespecified or pre-estimated tensor tubal-rank value. To achieve this, rank-1 basis tensors and weight tensors of the target tensor are estimated iteratively, and the residual error between the observed tensor and the estimated tensor through linear mapping is utilized as the stop condition. We theoretically prove the convergence and correctness of the proposed ORTP method. The methodology was tested on synthetic data, real video data, and video SAR data. These tests show that the proposed approach outperforms other video SAR imaging algorithms and low-rank tensor recovery algorithms.

The atmosphere affects the propagation of radar signals by provoking unwanted signal phase changes. In interferometric applications, such as coherent change detection and displacement measurements, this effect may significantly degrade the system performances. Moreover, atmosphere-induced phase changes are both time and space variants, and therefore, they are not easy to be removed. This article proposes a novel method to remove atmospheric effects by using a parametric model of the refractive index, which is derived as an extension of the International Telecommunication Union—Radiocommunication model. The proposed algorithm has been tested on real data acquired by using a ground-based synthetic aperture radar system in conjunction with data collected by a weather station. Data have been acquired continuously for three consecutive days, approximatively every 5 min. Results have shown how the proposed method can effectively remove atmospheric effects and restore the signal phase.

Wildfires are a major concern around the globe because of the immediate impact they have on people's lives, local ecosystems, and the environment. Soil moisture is one of the most important factors that influence wildfire occurrences and spread. However, it is also one of the most challenging hydrological variables to measure routinely and accurately. Therefore, soil moisture is significantly underutilized in operational wildfire risk applications. Thus, the aim here is to use a well-established operational soil moisture product to isolate the soil moisture-fire relationship and assess the utility of using soil moisture as a leading indicator of potential fire risk. We evaluated the value of remotely-sensed soil moisture observations from the soil moisture active passive sensor for monitoring and predicting fire risk in Australia and California. We quantified the relationship between observed fire activity and soil moisture conditions and analyzed the soil moisture conditions for two extreme fire events. Our findings show that fire activity is strongly associated with soil moisture anomalies. Lagged correlation analysis demonstrated that a remote-sensing based soil moisture product could predict fire activity with a 1–2 month lead-time. Soil moisture anomalies consistently decreased in the months preceding fire occurrence, often from normal to drier conditions, according to a spatiotemporal analysis of soil moisture in two extreme fire events. Overall, our findings indicate that soil moisture conditions prior to large wildfires can aid in their prediction and operational satellite-based soil moisture products such as the one used here have real value for supporting wildfire susceptibility and impacts.

Benefiting from free labeling pixel-level samples, weakly supervised semantic segmentation (WSSS) is making progress in automatically extracting building from high-resolution (HR) remote sensing (RS) imagery. For WSSS methods, generating high-quality pseudomasks is crucial for accurate building extraction.To improve the performance of generating pseudomasks by using image-level labels, this article proposes a weakly supervised building extraction method by combining adversarial climbing and gated convolution. The proposed method optimizes class activation maps (CAMs) by using adversarial climbing strategy, generates accurate class boundary maps by introducing a gated convolution module, and further refines building pseudomasks by fusing pairing semantic affinities and CAMs with a random walk strategy. Experimental results on three datasets—two ISPRS datasets and a self-annotated dataset—demonstrate that the proposed approach outperformed SOTA WSSS methods, leading to improvement of building extraction from HR RS imager. This article provides a new approach for optimizing pseudomasks generation, and a methodological reference for the applications of weakly supervised on RS images.

Due to technology and cost limitations, it is challenging to obtain high temporal and spatial resolution images from a single satellite spectrometer, which significantly limits the specific application of such remote sensing images in earth science. To solve the problem that the existing algorithms cannot effectively balance the spatial detail preservation and spectral change reconstruction, a pseudo-Siamese deep convolutional neural network (PDCNN) for spatiotemporal fusion is proposed in this article. The method proposes a pseudo-Siamese network framework model for fusion. This framework has two independent and equal feature extraction streams, but the weights are not shared. The two feature extraction streams process the image information at the previous and later moments and reconstruct the fine image of the corresponding time to fully extract the image information at different times. In the feature extraction stream, the multiscale mechanism and dilated convolution of flexible perception are designed, which can flexibly obtain feature image information and improve the model reconstruction accuracy. In addition, an attention mechanism is introduced to improve the weight of the crucial information for the remote sensing images. Adding a residual connection enhances the reuse of the initial feature information in shallow networks and reduces the loss of feature information in deep networks. Finally, the fine images obtained from the two feature extraction streams are weighted and fused to obtain the final predicted image. The subjective and objective results demonstrate that the PDCNN can effectively reconstruct the fusion image with higher quality.

Recently, remote sensing (RS) technology are becoming an increasingly important technology in estimating forest growing stem volume (GSV), and the saturation issue of spectral variables from various optical sensors severely hinders the improvement of mapping forest GSV, especially in the planted forest with high GSV. Forest canopy height is widely considered as one of the major factors to increase the saturation levels in mapping GSV. However, it is rather difficult to invert the forest canopy height without precisely external DEM for large regions. In this study, the canopy height model (CHM) was derived from ZY-3 stereo images with subtracting open-sourced external DEM and the response of saturation levels was analyzed by adding forest height in the planted coniferous forest (Larch and Chinese pine). To further describe the relationships between the forest height and saturation levels, five datasets with five estimation models (Linear, MLR, SVR, KNN, and RF) and three methods of variable selection (Stepwise, LASSO, and Pearson) were applied to estimate the forest GSV using corrected CHM and 49 alternative variables extracted from ZY-3 multispectral images. Meanwhile, a spherical model was employed to quantitatively describe the saturation levels of combined variables. The results showed that values of relative root means square error were decreased from 29.3% to 25% for Larch and from 26.5% to 22.2% for Chinese pine after adding the corrected CHM, respectively. Meanwhile, the saturation level of each combined variable set was successfully determined by the spherical model. The results illustrated that the saturation levels of GSV were significantly increased by adding corrected CHM from open-sourced external DEM. Specially, the averaged saturation levels were increased from 220 m3/ha to nearly 300 m3/ha for Chinese pine and from 150 m3/ha to 220 m3/ha for Larch, respectively. It is proved that ZY-3 stereo and multispectral images have great potential for accurate estimation of forest GSV by delaying the saturation levels using extracted CHM with open-sourced external DEM.

As two of the most widely used remote sensing images, optical and synthetic aperture radar (SAR) images show abundant and complementary information on the same target owing to their individual imaging mechanisms. Consequently, using optical and SAR images simultaneously can better describe the inherent features of the target, and thus, be beneficial for subsequent remote sensing applications. In this article, we propose a novel modular fully convolutional network model to improve the accuracy of land cover classification by fully exploiting the complementary features of the two sensors. We investigate where and how to fuse the two images in the joint classification network. A cross-gate module with a bidirectional information flow is proposed to achieve the best fusion performance. In addition, to validate the proposed model, we construct a multiclass land cover classification dataset. Exhaustive experiments show that the proposed joint classification network presents superior results than state-of-the-art classification models using single-sensor images.

Sparse unmixing with a semisupervised fashion has been applied to hyperspectral remote sensing imagery. However, the imprecise spatial contextual information, the lack of global feature and the high mutual coherences of a spectral library greatly limit the performance of sparse unmixing. In order to address these prominent problems, a new paradigm to characterize sparse hyperspectral unmixing is proposed, namely, the superpixel-based weighted collaborative sparse regression and reweighted low-rank representation unmixing (SBWCRLRU). In this method, the weighted collaborative sparse regression explores the pixels shared the same support set to help the sparsity of abundance fraction, and the reweighted low rank representation minimizes the rank of the abundance matrix to promote the spatial consistency of the image. Meanwhile, superpixel segmentation is adopted to cluster the pixels into different spatial homogeneous regions to further improve the unmixing performance. Extensive experiments results conducted on both synthetic and real data demonstrate the effectiveness of the proposed SBWCRLRU. It can not only improve the performance of hyperspectral unmixing but also outperform the existing sparse unmixing approaches.

Speckle removal process is inevitable in the restoration of synthetic aperture radar (SAR) images. Several variant methods have been proposed for enhancing SAR images over the past decades. However, in recent studies, convolutional neural networks (CNNs) have been widely applied in SAR image despeckling because of their versatility in representation learning. Nonetheless, a fair number of textures of the images are still lost when despeckling using simple CNN structures. To solve this problem, an encoder–decoder architecture was previously proposed. Although this architecture extracts features on different scales and has been shown to yield state-of-the-art performance, it still learns representation locally, resulting in missing overall information of convolutional features. Therefore, we herein introduce a new method for SAR image despeckling (SAR-CAM), which improves the performance of an encoder–decoder CNN architecture by using various attention modules. Moreover, a context block is introduced at the minimum scale to capture multiscale information. The model is trained via a data-driven approach using the gradient descent algorithm with a combination of modified despeckling gain and total variation loss function. Experiments performed on simulated and real SAR data demonstrate that the proposed method achieves significant improvements over state-of-the-art methodologies.

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.

Sea ice concentration algorithms using brightness temperatures ($T_{B}$) from satellite microwave radiometers are used to compute sea ice concentration ($c_{\text{ice}}$), sea ice extent, and generate sea ice climate data records. Therefore, it is important to minimize the sensitivity of $c_{\text{ice}}$ estimates to geophysical noise caused by snow/sea ice thermal microwave emission signature variations, and presence of WV and clouds in the atmosphere and/or near-surface winds. In this study, we investigate the effect of geophysical noise leading to systematic $c_{\text{ice}}$ biases and affecting $c_{\text{ice}}$ standard deviations (STD) using simulated top of the atmosphere $T_{B}$s over open water and 100% sea ice. We consider three case studies for the Arctic and the Antarctic and eight different $c_{\text{ice}}$ algorithms, representing different families of algorithms based on the selection of channels and methodologies. Our simulations show that, over open water and low $c_{\text{ice}}$, algorithms using gradients between V-polarized 19-GHz and 37-GHz $T_{B}$s show the lowest sensitivity to the geophysical noise, while the algorithms exclusively using near-90-GHz channels have by far the highest sensitivity. Over sea ice, the atmosphere plays a much smaller role than over open water, and the $c_{\text{ice}}$ STD for all algorithms is smaller than over open water. The hybrid and low-frequency (6 GHz) algorithms have the lowest sensitivity to noise over sea ice, while the polarization type of algorithms has the highest noise levels.

The state-of-the-art crop phenological classifiers use vegetation index (VI) time-series data and deep learning (DL) techniques. However, the scarcity of training samples limits the performance of these approaches. Unlike the conventional augmentation techniques, the data augmentation of VI curves should preserve the crop-specific phenological events. The DL-based augmentation approaches do not give good results when the training samples are limited. Also, the conventional approaches such as translation, rotation, scaling, and wrapping do not preserve the characteristic features of the index curves, thereby making them inappropriate for the VI-curve-based augmentations. This article proposes a non-DL-based data augmentation strategy that requires only a minimal number of actual training samples. In the proposed approach, the periodic phenological events and the underlying trend for each crop class are modeled to improve the augmentation. The trends of different crop classes are estimated by jointly maximizing the autocorrelation and variance, while the optimal subsequences are generalized as the phenological events. The proposed augmentation strategy of using Maximal overlap discrete wavelet transform for obtaining the surrogates that retain the crop-specific features and periodicities significantly improves the results. It may be noted that the proposed approach does not alter the wavelet coefficients that are characteristics of a given crop class. The experiments using time series VI data, covering 90 fields of wheat, and 60 fields of barley, confirm better accuracy of the proposed augmentation approaches as compared to the prominent approaches.

This article presents a study on the power consumption and performance analysis of a small, portable, and ultra-low power total power L-band radiometer. The article explores two processing architectures: the ZYNQ 7010 and the ARM A8 embedded microprocessor. The processing algorithm based in C++ was tested for different clock frequencies, ADC sampling speeds, and sizes of the ADC buffer. To reduce the power consumption and the algorithm execution time, high-level and system-level optimizations, along with fixed-point Q(16,16) data representation, were applied to the main code running on LINUX Debian V8. In the case with the ZYNQ 7010, the optimizations had no notable impact on reducing power or execution time in comparison with the ARM A8, where significant variations were measured, showing a tradeoff between power consumption and algorithm performance that limits the processing capability and the system flight time. The ZYNQ 7010 runs the algorithm faster, but the power consumption is higher than the ARM A8. Using the fixed-point Q(16,16) implementation reduced the power consumption and the execution time in both architectures. Based on these results, we developed a heuristic methodology to minimize power consumption and increase the performance. Energy consumption savings for the radiometer during 20 min of flight was 48%. The size of the radiometer was reduced to 30 cm × 30 cm × 10 cm, with a weight of 1.36 kg, (3 lb) allowing the system be carried by small drones. The results were validated measuring salinity at two locations in Western Puerto Rico.

Airborne laser scanning point cloud semantic labeling, which aims to identify the category of each point, plays a significant role in many applications, such as forest observing, powerline extraction, etc. Under the guidance of deep learning technology, the interpretation thought of point clouds has also greatly changed. However, owing to the irregular and unordered natures of point clouds, it is relatively difficult for classification model to distinguish some objects with similar geometry by single-modal data only. Fortunately, additional gain information, e.g., color spectrum which can be complementary to geometric information, is able to effectively promote the classification effect. Therefore, the design of fusion strategy is a critical part in model construction. In this article, aiming to capture more abstract semantic information for color spectrum data, we elaborate a color spectrum fusion (CSF) module. It can be flexibly integrated into a classification pipeline with just negligible parameters. Then, we expand data fusion thoughts for point clouds and color spectrum and investigate three possible fusion strategies. Accordingly, we develop three architectures to construct CSF-Nets. Ultimately, by taking a weighted cross entropy loss, we can train our CSF-Nets in an end-to-end manner. Experiments on two extensively used datasets: Vaihingen 3D and LASDU show that the presented three fusion approaches all can improve the performance, while the earlier fusion strategy performs the best. Besides, compared with other well-performed methods, CSF-Net is still able to achieve satisfactory performance on overall accuracy and m$F_{1}$-score indicator. This also validates the effectiveness of our multimodal fusion network.

Compared with land surface temperature (LST) and land surface emissivity (LSE) retrieval from single-band or multispectral thermal infrared (TIR) data, TIR hyperspectral imagery allows us to obtain accurate LST and LSE through the use of an automatic temperature and emissivity separation (TES) method. However, the existing TES algorithms have rarely been investigated with airborne TIR hyperspectral imagery from the Hyper-Cam sensor, which is based on Fourier transform infrared technology. In this article, a practical LST and LSE retrieval framework incorporating reanalysis atmospheric profiles is proposed for use with Hyper-Cam airborne TIR hyperspectral imagery. In this framework, an atmospheric compensation method is introduced based on spatiotemporal analysis and the fusion of three types of widely used atmospheric profiles to replace the unattainable synchronously measured atmospheric profiles. An empirically constrained TES method is then proposed to extend the original TES algorithm to Hyper-Cam hyperspectral imagery. In addition, to exclude the negative effects of radiometric calibration error, measurement noise, and the atmospheric absorption lines in certain bands, the problematic bands are removed to improve the data quality. To evaluate the performance of the proposed framework, a set of airborne TIR hyperspectral imagery with 81 bands was acquired using the Hyper-Cam airborne system. Experiments were carried to compare the performance of the proposed method and the Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes-Infrared (FLAASH-IR) method. The results indicate that the proposed method can obtain more robust and accurate results than FLAASH-IR, and the root-mean-square error (RMSE) of the emissivity data is around 0.015 for all the validation samples.

Precise ionospheric total electron content (TEC) is critical for many aerospace applications, and forecasting ionospheric TEC is of great significance to it. Besides, short-term prediction of TEC values fills the gap between the TEC product latency and the precision. The machine learning-based approaches are promising in solving the nonlinear prediction issues, particularly suitable for short-term global positioning system TEC forecasting due to its complex temporal and spatial variation. In this article, four different machine learning models, i.e., artificial neural network, long short-term memory networks, adaptive neuro-fuzzy inference system based on subtractive clustering, and gradient boosting decision tree (GBDT) are applied for forecasting ionospheric TEC in three IGS GNSS monitoring stations at the low-latitude region (16°S to 10°S). The performance of these approaches in extreme conditions is investigated, including the high solar activity and magnetic storm, which are the most challenging scenario for TEC prediction. The results show that the machine learning algorithms outperform the global ionospheric map prediction model. The prediction accuracy during the high solar activity period was improved from 37.93% to 49.28%. During the magnetic storm period, the prediction accuracy was improved from 28.16% to 67.39%. Among the machine learning algorithms, the GBDT model outperforms the rest three algorithms in ionosphere prediction scenarios, which improves the prediction accuracy by 5.6% and 12.7% than the rest three approaches on average during high solar activity (2012–2015) and magnetic storm periods respectively.

Multiimage super-resolution (MISR), as one of the most promising directions in remote sensing, has become a needy technique in the satellite market. A sequence of images collected by satellites often has plenty of views and a long time span, so integrating multiple low-resolution views into a high-resolution image with details emerges as a challenging problem. However, most MISR methods based on deep learning cannot make full use of multiple images. Their fusion modules are incapable of adapting to an image sequence with weak temporal correlations well. To cope with these problems, we propose a novel end-to-end framework called TR-MISR. It consists of three parts: An encoder based on residual blocks, a transformer-based fusion module, and a decoder based on subpixel convolution. Specifically, by rearranging multiple feature maps into vectors, the fusion module can assign dynamic attention to the same area of different satellite images simultaneously. In addition, TR-MISR adopts an additional learnable embedding vector that fuses these vectors to restore the details to the greatest extent. TR-MISR has successfully applied the transformer to MISR tasks for the first time, notably reducing the difficulty of training the transformer by ignoring the spatial relations of image patches. Extensive experiments performed on the PROBA-V Kelvin dataset demonstrate the superiority of the proposed model that provides an effective method for transformers in other low-level vision tasks.

Image matching between the optical and synthetic aperture radar (SAR) is one of the most fundamental problems for earth observation. In recent years, many researchers have used hand-made descriptors with their expertise to find matches between optical and SAR images. However, due to the large nonlinear radiation difference between optical images and SAR images, the image matching becomes very difficult. To deal with the problems, the article proposes an efficient feature matching and position matching algorithm (MatchosNet) based on local deep feature descriptor. First, A new dataset is presented by collecting a large number of corresponding SAR images and optical images. Then a deep convolutional network with dense blocks and cross stage partial networks is designed to generate deep feature descriptors. Next, the hard L2 loss function and ARCpatch loss function are designed to improve matching effect. In addition, on the basis of feature matching, the two-dimensional (2-D) Gaussian function voting algorithm is designed to further match the position of optical images and SAR images of different sizes. Finally, a large number of quantitative experiments show that MatchosNet has a excellent matching effect in feature matching and position matching. The code will be released at: https://github.com/LiaoYun0x0/Feature-Matching-and-Position-Matching-between-Optical-and-SAR.

Since 1958, the Mauna Loa Observatory (MLO) has continuously monitored carbon dioxide variations using nondispersive infrared sensors, with the Keeling curve as an early indicator of the anthropogenic contribution to global atmospheric carbon dioxide. The increasing CO2 levels are alarming and have led to international agreements that promote cleaner industrial activities. However, any change in global behavior would not immediately cause detectable changes in the MLO data; the extent to which global and long-term trends are conflated with local and short-term variations remains unclear. Hence the current article verifies the performance of the sampling and measurement systems of MLO, using existing data published within the months of October and November 2020, which comply with the temporal continuity requirements of chronostatistics. It has been determined that the components of the MLO air including carbon dioxide are well mixed due to their particular location. Beyond this, the variographic analysis distinguishes between small (10%) variability contributions due to sampling, including graphical depictions of MLO data. Coupled with the precision of the method being better than 0.2 ppm, it has been determined that the sampling and measurement protocols are highly suitable to meet the objective of representing CO2 fluctuations over time. The variographic application also manages to quantify short-term variabilities resulting from the local processes of the region where the observatory is located. The results support the furthering of multiscaled temporal analysis of atmospheric CO2, and potentially the incorporation of CO2 variographic parameters into empirical and semiempirical climate models.

Relationship between urban diversity and urban vitality is imperative for guiding better design in urban development, although existing frameworks are not able to efficiently examine the relationship at multiple scales. In this article, we propose a new framework to integrate nighttime light (NTL) imagery and multisource urban data into multiscale geographically weighted regression (MGWR) models to examine the varying relationship between diversity and vitality across space and time. NTL is used as a proxy for urban nighttime vitality. Public transport, taxi transit, and points of interest data are used to derive three aspects of urban diversity indices: ridership diversity, spatial interaction diversity, and built environment diversity. By comparing the models in holiday and nonholiday weeks in Shenzhen, China, the NTL-based vitality proxy was found to be strongly correlated with the urban diversity indices, given by the satisfactory goodness of fit (r-squared = 0.9) of the MGWR models. The spatially varying relationships between diversity indices and nighttime vitality were observed and patterns discussed. The analysis of the coefficients revealed the importance of stable public transport and fluctuating taxi trips for nighttime vitality. The new index proposed for the diversity of spatial interaction (DSI) is a strong indicator for nighttime vitality, adding to existing vitality indicators. Furthermore, this study found that DSI and density of catering have less temporal variation, indicating their robustness in measuring nighttime vitality. This study provided empirical insights into how nighttime vitality is related to urban diversity, demonstrating new applications of NTL for intracity studies.

Sea surface height (SSH) retrieval based on spaceborne Global Navigation Satellite System Reflectometry (GNSS-R) usually uses the GNSS-R geometric principle and delay-Doppler map (DDM). The traditional method condenses the DDM information into a single scalar measure and requires error model correction. In this article, the idea of using machine learning methods to retrieve SSH is proposed. Specifically, two widely-used methods, principal component analysis combined with support vector regression (PCA-SVR) and convolution neural network (CNN), are used for verification and comparative analysis based on the observation data provided by Techdemosat-1 (TDS-1). According to the DDM inversion method, ten features from TDS-1 Level 1 data are selected as inputs; The SSH verification model based on the Danmarks Tekniske Universitet (DTU) 15 ocean wide mean SSH model and the DTU global ocean tide model is used as output verification of SSH. For the hyperparameters in the machine learning model, a grid search strategy is used to find the optimal values. By analyzing the TDS-1 data from 31 GPS satellites, the mean absolute error (MAE), root-mean-square error (RMSE) and coefficient of determination (R2) of the PCA-SVR inversion model are 0.61 m, 1.72 m, and 99.56%, respectively; and the MAE, RMSE, and R2 of the CNN inversion model is 0.71 m, 1.27 m, and 99.76%, respectively. In addition, the time required to train the PCA-SVR and CNN inversion models is also analyzed. Overall, the technique proposed in this article can be confidently applied to SSH inversion based on TDS-1 data.

This study aims to introduce effective unmanned aerial system (UAS) survey conditions for coastal sediment classification, including muddy sand, sand, gravel, and shells in a tidal flat area. UAS images with resolutions ranging from 2 to 60 mm are used as an implication of survey altitudes. The UAS images are used for sediment classification using random forest (RF) and support vector machine (SVM) methods. The results showed that RF is more effective in sediment classification while the general accuracy pattern was similar. The accuracy decreased with lower spatial resolutions. Notably, there is a significant drop of accuracy with a resolution coarser than 40 mm. Considering the training data selection, classification accuracy, and survey efficiency, it is suggested that 40 mm UAS images would provide optimal condition with acceptable accuracy for coastal sediment classification using RF model. To gain higher accuracy, a lower flight altitude is required, which will elongate the survey time significantly. Given the fact that this study is the first approach to test various UAS survey conditions for coastal sediment classifications in a field condition; the methodology and findings of this study can serve as a guideline framework for future coastal UAS sediment mapping.

Understanding the variation regularity of water extent can provide insights into lake conservation and management. In this study, inter- and inner-annual variations of water extent during the period of 1987–2020 were analyzed to understand the temporal and spatial distribution characteristics of Dongting Lake. We applied the Multiple Index Water Detection Rule to extract the Dongting Lake water extent quickly and accurately based on Google Earth Engine platform, and then assessed the extraction accuracy. The water surface analysis results showed that (1) based on sustainable development goals (SDG) 6.6.1, the trend of water extent showed the downward fluctuating trend from 1987 to 2020, with the overall average water extent being 1894.48 km². (2) Among the monthly average water area, the largest extent was 2477.14km² (July) and the smallest was 848.14 km² (January). Among the seasonal mean water area, summer was the largest, with an area of 2438.06 km², and winter was the smallest at 967.34 km². (3) For the water inundation frequency, seasonal water bodies accounted for the largest proportion, with 1577.85 km²; the nonwater area was the smallest, with the area of 573.02 km²; and the permanent water area was 1086.21 km². Through the analysis of the historical water body extent of the long time series of Dongting Lake, this study reflected support for SDG, for which the research idea and design can help us understand the importance and feasibility of the SDG 6.6.1 indicator.

The polar ice sheets have undergone unprecedented melt events in the recent past, which have consequences for ice sheet mass balance, stability, and sea level. In this study, we employed L-band (1.4 GHz) brightness temperature observations collected by NASA's Soil Moisture Active Passive (SMAP) mission to investigate the extent, duration and intensity of melt events on the Antarctic Ice Sheet from 2015 to 2020. Satellite microwave measurements have long been used to detect melt events because of their sensitivity to the presence of liquid water in snow and ice. The observed microwave response depends on the sensor measurement frequency. Our hypothesis for this study is that the relatively long wavelength SMAP observations can detect a wider range of surface wetness conditions relative to shorter wavelength microwave observations that attain signal saturation at relatively lower wetness levels and within shallower surface layers. SMAP provides nearly all-weather surface monitoring over all of Antarctica twice daily with morning and evening overpasses at about 40 km spatial resolution. We applied an empirical threshold algorithm using horizontally and vertically polarized microwave brightness temperature differences to detect surface melt events over Antarctica from 2015 through 2020. The results show that the SMAP empirical algorithm can be used to detect melt extent and duration, and the geophysical model-based algorithm can be used to detect snow wetness, which can be used as an indicator of melt intensity. Analysis of the melt seasons between 2015 and 2020 show that the even though the melt extent in 2019–2020 was not as large as during the 2015–2016 melt season, it was significantly more intense, particularity on the West Antarctic Ice Sheet.

In this article, we propose an adaptive superpixel generation algorithm for synthetic aperture radar (SAR) imagery, which is implemented based on fast density-based spatial clustering of applications with noise (DBSCAN) clustering and superpixel merging with edge penalty. The superpixel generation algorithm consists of two stages, i.e., fast pixel clustering and superpixel merging. In the clustering stage, we define a new adaptive pixel dissimilarity measure for SAR image and then optimize the DBSCAN strategy, which considers the edge information and can achieve rapid clustering. In the merging stage, based on the initial superpixels, a new superpixel dissimilarity measure is defined, which can merge the small local superpixels into their neighborhood superpixels, making the final superpixel segmentation results compact and regular. Experimental results on two simulated and two real SAR images demonstrate that our method outperforms the state-of-the-art superpixel generation methods in terms of both efficiency and accuracy. The superpixel segmentation accuracy of our method is 5–10% higher and the time cost is 10–40% lower than other methods. Since the superpixel segmentation result can be used as a preprocessing stage for the SAR data interpretation applications, superpixel-based and pixel-based classification results with two real SAR images are also used for comparison, which can validate the advantages of our proposed method.

Deep learning (DL) has heavily impacted the data-intensive field of remote sensing. Autoencoders are a type of DL methods that have been found to be powerful for blind hyperspectral unmixing (HU). HU is the process of resolving the measured spectrum of a pixel into a combination of a set of spectral signatures called endmembers and simultaneously determining their fractional abundances in the pixel. This article details the various autoencoder architectures used in HU and provides a critical comparison of some of the existing published blind unmixing methods based on autoencoders. Eleven different autoencoder methods and one traditional method will be compared in blind unmixing experiments using four real datasets and four synthetic datasets with different spectral variability. Additionally, extensive ablation experiments with a simple spectral unmixing autoencoder will be performed. The results are interpreted in terms of the various implementation details, and the question of why autoencoder methods are so powerful compared to traditional methods is unraveled. The source codes for all methods implemented in this article can be found at https://github.com/burknipalsson/hu_autoencoders.

This article investigates the potential of the assimilation of both synthetic aperture radar (SAR)-derived dry mass (DM) and optically derived green area index (GAI) in an agro-meteorological model (i.e., SAFY) for a better rapeseed crops modeling. The GAI was derived from both 566 Sentinel-2 and 149 Landsat-8 images, whereas DM was derived from 884 Sentinel-1 images acquired from six different orbits. The ground data were collected during 3 agricultural years on 43 rapeseed fields located in three study areas in France with contrasted pedoclimatic conditions. Results show that the temporal evolutions of both DM and GAI can be accurately simulated over the 43 monitored rapeseed fields (R2 = 0.84 and 0.92, respectively, and relative root mean square error, RMSEr = 41% and 28%, respectively) for the best satellite configuration. Tested assimilation scenarios reveal that the concomitant assimilation of SAR and optical data allows a significant better control of the model than the assimilation of SAR or optical data alone. Small differences in the simulations of the model are observed when it is controlled by either multisensor optical and/or multiorbital SAR data or monosensor optical and/or mono-orbital SAR data. The discrepancies of performance between fields push toward the strengthening of this study by considering other rapeseed fields with ground observations acquired worldwide for the calibration of SAR-derived DM. These results are however promising in view of the development of a near-real time assimilation scheme as a decision support tool for farmers and decision makers.

A geostationary (GEO) hyperspectral infrared sounder (HyIRS) is capable of providing high spectral (0.625 cm−1), temporal (every 30 min) and spatial (4 km) resolution observations over the continental U.S. (CONUS). Frequent observations from a GEO-HyIRS at high spatial resolution are expected to contribute to the generation of three-dimensional structures of atmospheric temperature and humidity, and wind. These new observations will provide valuable information for timely forecasts of severe storms over the CONUS and the overall Western Hemisphere. Infrared (IR) sounder observations from a geostationary orbit open a new set of possibilities, including the capability of monitoring the diurnal cycle of atmospheric patterns, which is difficult from Low Earth Orbit IR sounders and the capability of timely and accurate retrievals of several trace gases. In this article, the feasibility of adding a HyIRS into the next generation of U.S. geostationary environmental satellites is studied. The configuration of a notional U.S. GEO-HyIRS sensor and its ground data processing system are discussed. A hyperspectral IR data simulator is developed and reported as part of this engineering study, where proxy data is used to model the end-to-end ground processing system. Various considerations for the configuration and the calibration and validation of the instrument are addressed.

Over the past decade, many excellent data sharing efforts have enriched the remote sensing scene classification (SC) methods. These datasets have achieved great success in complex high-level semantic information interpretation. However, most existing datasets are collected from standard and ungeoreferenced image patches for algorithm training and evaluation. These datasets do not fit for practical applications and cannot be directly applied in further geographical study. Accordingly, we provide a large range high-resolution SC dataset with multiple time phases, called “Wuhan Multiapplication VHR Scene classification dataset (WH-MAVS).” It facilitates the study of SC and scene change detection (SCD) algorithms. Moreover, it can also be directly employed to perform a variety of real-life land use application tasks. To the best of our knowledge, this is the first free, publicly available, georeferenced, and annotated dataset to cover almost an entire megacity. The WH-MAVS was collected and annotated from Google Earth imagery with the same spatial resolution and uniform nonoverlapping patch size, covering the central area of Wuhan, China. The total number of scene samples is 47 137, which belong to 14 classes with 23 567 labeled patches for each time phase in 2014 and 2016, respectively. The geographic coordinates of all samples in both time phases exhibit one-to-one correspondence with 23 202 unchanged image patches of scene categories and 365 changed ones. The distribution of the number of samples in each class is highly imbalanced; moreover, there are large intraclass differences and indistinguishable interclass variances. These characteristics are closer to the real land use/land cover application tasks and introduce further challenges to the related algorithm research. In addition, we conducted benchmark experiments on SC and SCD based on the WH-MAVS dataset with widely used deep learning models. DenseNet169 was found to achieve the best performance. The overall accuracies are 91.07% and 92.09%, respectively, in the 2014 and 2016 validation sets of WH-MAVS. Furthermore, SCD obtained by DenseNet169 has a binary change detection accuracy of 89.56% and a multiple (from–to) change detection accuracy of 86.70%. Over and above the research value of the algorithm, it is also proven to have practical applications in fields such as urban planning, landscape pattern analysis, and urban dynamic monitoring and analysis.

Enhancement of vessel detection performance in synthetic aperture radar (SAR) images generated academic advancements related to amelioration of the algorithmic accuracy and training data procurement. For practical implementation of vessel detection algorithm to maritime surveillance, however, presentation of authentic position of vessels was essential. Accordingly, this article aimed to propose an algorithm, which demonstrated realistic and azimuth shift-corrected position of vessel, especially out of conventional vessel monitoring apparatus: automated identification system (AIS) and vessel-pass (VPASS) information. Two different analyses regarding the vessel detection output utilization were, therefore, presented. Primary analysis demonstrated a vessel identification algorithm, comparing the vessel detection output with elaborately preprocessed AIS and VPASS information, which indicated the discrete position and velocity of vessel. The other presented a position restoration algorithm via i) velocity estimator complementing the conventional fractional Fourier transform velocity estimation analysis, while investigating the effect of range acceleration in deriving the azimuth velocity and ii) measuring the vessel orientation angle from Radon transform. Both algorithms were implemented to the vessel detection output in Cosmo-SkyMed SAR images, depicting an enhanced accuracy compared to the conventional algorithm both in velocity estimation and azimuth shift estimation; velocity offset reduced from 1.64 m/s to 1.29 m/s and average azimuth shift offset reduced from 70.75 m to 62.39 m. The presented algorithms would be decisive in terms of practicality if robustly attached to convolutional neural network-based vessel detection algorithms demonstrating ideal detection performances.

Convolutional neural networks (CNNs) have significance in remote sensing image mapping, and pixel-level representation allows refined results. Due to inconsistencies within a class and different scales of water bodies, the water body mapping has challenges, such as insufficient integrity and rough shape segmentation. To resolve these issues, we proposed an intelligent water bodies extraction method (named SADA-Net) for high-resolution remote sensing images. This method considers multiscale information, context dependence, and shape features. The network framework integrates three critical components: shape feature optimization (SFO), atrous spatial pyramid pooling, and dual attention modules. SADA-Net can accurately extract an extensive range of water bodies in complex scenarios. SADA-Net has certain advantages regarding small and dense water bodies extraction, as the SFO module effectively solves the defects of the unified processing of low-level features in the encoder stage of CNNs, which highlights the shape information of a water body. Two data types (red, green, and blue bands and multispectral images) are employed to verify the performance of the proposed network. The best result achieved an evaluation index F1-Score of 96.14% in large-scale image segmentation, and the structural similarity index measure reached 94.70%. Overall, the proposed method achieves the purpose of maximizing the integrity and optimizing the shape of a water body. Additionally, the SADA-Net proposed in this article has a specific reference value for high-resolution remote sensing image water bodies mapping.

In this article, we develop a novel fully unsupervised autoencoder-based scheme for nonlinear hyperspectral pixel unmixing. A unique approach is derived where high noise and unresponsive pixels are accounted for, by a unique averaging approach based on spatially aware filters built using radial basis function (RBF) kernels. A novel technique is implemented via calculating rank-equivalent kernel covariance matrices in order to estimate the unknown number of endmembers contributing to the data. Utilization of spatial information is done via RBF-based weighted averaging, which is then followed by endmember estimation via K-means clustering. The RBF distances from the cluster centers are determined to measure the position of the mixed pixels in relation to the centers, which is utilized as a preliminary estimation of the abundances. The proposed framework is robust in the presence of unresponsive pixels, while highly versatile working with different nonlinear unmixing models. Extensive numerical tests establish the superiority of the novel approach with respect to state-of-the-art methods.

Subspace clustering methods have become a powerful tool to cluster hyperspectral imaging (HSI) data as they ensure theoretical guarantees and empirical success. However, existing methods explore subspace representation in the Euclidean space, and thus, failing to exploit the high-order relationship and long-range interdependences. This article presents a simple yet effective method, to extend subspace clustering into the non-Euclidean domain entitled hypergraph convolutional subspace clustering (HGCSC). Instead of treating HSI as Euclidean data only, we represent all the intraclass relations as hyperedges in a hypergraph. With this representation, we can recast the classic self-expression as a hypergraph convolutional self-representation model. To explore the long-range neighboring relation, we introduce a multihop hypergraph convolution process into the method by collapsing the repeated multiplications into a single matrix. HGCSC adopts the Frobenius norm to ensure a closed-form solution. We assess the performance of HGCSC on five real HSI datasets and show that HGCSC significantly outperforms competitors in terms of clustering accuracy.

In recent times, multisource remote sensing technology [e.g., hyperspectral image (HSI) and light detection and ranging (LiDAR) data] has been widely used in urban land-use recognition owing to its high classification effectiveness compared to using only single-source data. In this study, a multiview hierarchical network (MVHN) technique is developed for HSI and LiDAR data classification, which conducts the following execution procedures. First, based on the a preset band step length, the original HSI is sampled and divided into multiple groups with exactly the same number of bands to obtain spectral features. Then, principal components analysis is performed on the raw HSI to extract the first principal components (PCs) that meet the size of the LiDAR image. The Gabor filters are applied to the PCs and LiDAR to capture spatial details (i.e., textural features) of scenes. Specifically, a stacking mechanism is employed to generate fusion features once the above features are available. Next, a three-dimensional ResNet-like deep CNN is designed to extract the spectral–spatial information of the fusion feature. Finally, majority-voting is introduced into the classification results of the network trained using each fusion feature to achieve high-confidence final results. Experiments on three well-known HSI and LiDAR datasets (i.e., Houston, MUUFL, and Trento datasets) demonstrate the effectiveness of the proposed MVHN method compared to state-of-the-art comparable classification methods.