Oil palm phenology has many advantages in managing the sustainability of oil palm plantations. The phenology of oil palms is a key issue in harvest estimation, fruit bunch production, estimating oil palm taxes, replanting, fertilization, and detecting oil palm disease. One of the recently developed methods of oil palm phenology involves the use of remote sensing technology. We evaluated and reviewed the current state of oil palm phenology based on remote sensing and conducted an optimized systematic review of recent scientific publications, specifically focusing on scientific peer-reviewed papers published between 1990 and 2021, comprising over 100 existing journal papers on remote sensing for oil palm phenology. The review includes a description of the state of the art and the mapping of oil palm phenology based on sensors, biophysical tree parameters, and classification techniques and also describes the state of the art in the development of regression models of oil palm phenology based on wavelength, biophysical tree parameters, and the type of regression model. Finally, the review provided an opportunity to develop suitable techniques for the identification, classification, and the construction of regression models of oil palm phenology. There is a lot of potential in combining multisensor approaches, suitable classification methods, and regression models for oil palm phenology. For future studies on oil palm phenology, we recommend integrating machine learning with oil palm biophysical parameters based on multisensor remote sensing technologies.

Guest Editors Panagiotis Partsinevelos and Hongbo Su introduce the Special Section on Unmanned Systems and Satellites: A Synergy for Added-Value Possibilities.

Accurate and timely crop yield estimation prior to harvest is important for agricultural management, agricultural economy, and food security. In many cases, the farmers estimate the yield visually. Further, several crop simulation models have been developed to estimate yield accurately. However, these are not used efficiently because of their requirements for enormous amounts of data and their inability to show the spatial differences of yield in the field. Recently, the rapid development of unmanned air vehicle (UAV) technologies has shown great potential to estimate crop yield accurately and show the spatial heterogeneity in the field. We estimate the bell pepper yield with time series, high-resolution UAV multispectral images. To do so, canopy volume and five different spectral vegetation indices used widely were calculated. Seven UAV flight missions were conducted between June and August of 2019. Various linear regression models were developed to estimate the bell pepper yield based on the canopy volume values and vegetation indices. The results showed that the bell pepper canopy volume fit the yield best with the minimum estimation error [coefficient of determination   (  R²  )    =  0.93 and root mean square error   (  RMSE  )    =  2.30  tons ha^−  1]. In addition, a significant correlation was found between the enhanced vegetation index and bell pepper yield (R²  =  0.87 and RMSE  =  3.16  tons ha^−  1).

Remotely piloted aircraft systems (RPAS) have introduced a new ability to quickly deploy low-cost, fully or partially autonomous aerial sensor platforms, which has created new intelligence, surveillance, and reconnaissance capabilities in various domains using cameras that are ubiquitous in most RPAS. The mounted cameras acquire images or full-motion video (FMV) which can be analyzed using object detection algorithms for locating and classifying one or more specified targets. To date, there has not been much published work regarding the effect of the RPAS flight parameters on the performance of object detection algorithms. To explore the use of object detection on aerial FMV acquired at various RPAS flight parameter settings, a dataset acquisition campaign was launched resulting in 8.5 h of RPAS-acquired FMV. Analysis and interpretation of the acquired dataset revealed that state-of-the-art performance was achieved using a modified you only look once object detection algorithm when the RPAS was deployed under an altitude of 30 m, at a velocity of under 7 m/s, and at pitch angles ranging from 25 deg to 65 deg while acquiring FMV at a resolution of 4.16 MP. The experimental results show that, when flown under specific conditions, RPAS are an effective and reliable platform for acquiring aerial FMV for the purpose of object detection which has a variety of different applications, such as peace support, public safety, and aerial monitoring.

Crop evapotranspiration (ET) is a crucial component of energy and water budgets. The accurate determination of ET is vital for agricultural water management. Several satellite-based ET models have been developed to map ET at the field to regional scales. The spatial resolution of the satellite observations, particularly thermal-infrared imagery, is insufficient to estimate ET for small (<1  ha) agricultural fields. With unmanned aerial vehicle (UAV) technology advancement, high spatial and temporal images can be acquired with UAVs to monitor ET for small fields or even at a canopy scale. The aim of this study was the evaluation of the two-source energy balance (TSEB) model to estimate daily and seasonal crop (bell pepper) ET (ET_TSEB) using high-resolution visible and thermal UAV imagery. Also, the impact of using different pixel resolutions (40, 50, 60, and 70 cm) with TSEB is compared with ET values derived using a soil water budget approach (ET_SWD) with a profile soil water content. The results of this study showed that there is a high correlation between ET_TSEB and ET_SWD values (R²  =  0.73 for daily, R²  =  0.98 for seasonal). The root mean square error values for daily and seasonal ET_TSEB are 0.62  mm day^−  1 and 11.46  mm season^−  1, respectively. The sensitivity of TSEB output to the spatial resolution indicated that different pixel resolutions do not significantly impact ET estimates. We suggest that the TSEB model has a real potential for agricultural water management applications for small agricultural fields using high-resolution UAV multispectral and thermal images.

A valuable archive of multiresolution imagery covering the period from 1960 to 1980 has been declassified by the United States of America (USA) government on February 22, 1995, and became available for sharing to the scientific community. This imagery is part of USA’s top-secret program with the codename “CORONA” that ran since the Cold War. We tried to assess the accuracy of the declassified datasets and estimate if they can be used in synergy with classical aerial photos to estimate the future shoreline position. Furthermore, their contribution to estimating the shoreline’s long-term changes using statistical tools such as the linear regression rate was examined. Two elongated coasts located in the Pinios and Andravida/Kyllini Municipalities located in the north of Ilias Prefecture, Peloponnese, Greece, were used as test sites. Two models were used that compared the actual and the forecasted. The results show that erosion prevails in the coastal area of the Pinios and Andravida/Kyllini municipalities with a mean EPR rate ranging from −0.16 to −0.43  m  /  y and from −0.18 to −0.31  m  /  y according to the actual and the forecasted models, respectively. Moreover, the respective maximum erosion rates are −1.15 and −1.57  m  /  y (actual model) and −2.04 and −1.89  m  /  y (forecasted model), whereas the maximum accretion rates are +0.76 and +1.02  m  /  y (actual model) and +1.68 and +2.00  m  /  y. The declination of the CORONA declassified imagery ranges from 2.63% to 6.62% in the Andravida/Kyllini territory and from 6.53% to 22.38% in the Pinios area. These results were cross-validated, using the R-squared rates, the root mean square error, and the mean annual error (MAE) statistical techniques. The MAE presented a better correlation as it provided equal rates (0.31  m  /  y) for both of the EPR datasets.

Urban forestry is regarded as the green infrastructure of cities, the analysis of which is of great significance for research on statistical analysis, such as greening, carbon sink, etc. on the basis of remote sensing detection. However, current urban forestry detection algorithms are focused on typical machine learning, whose results are limited by the manually selected features. Therefore, an innovative detection algorithm for analyzing urban forestry is proposed. First, the simple linear iterative clustering algorithm is adopted to segment the dataset, and the edge information of the object is preserved as much as possible. Second, a new network structure is constructed on the basis of Mask-RCNN to refine the detection results: (1) dilated convolution is adopted to replace the normal convolution of C1 in backbone (ResNet101) to increase the size of receptive field; (2) convolutional block attention module is introduced in C2 to give more weight to the shallow features such as edges and colors; and (3) atrous spatial pyramid pooling (ASPP) is added to branch of C3, enhancing the ability of multiscale fusion. With the effect of ASPP on detection results at different positions of backbone compared, the results show that the more backward the ASPP is, the more obvious the trend of negative correlation between network convergence speed and detection accuracy (C3: 89.2%, C4: 84.3%, and C5: 81.6%) is. The accuracy of random forest and support vector machines is also compared, with the former of 69.4% and the latter of 71.2%. In addition, the method of transfer learning to unmanned aerial vehicle data is used in this research, with the accuracy of 81.6%. The result shows that the improved segmentation model features better ability of learning shallow features than the original, able to accurately extract urban forestry from remote sensing images, which provides the reference for the further research on urban forestry detection, mapping, as well as the multisource remote sensing data fusion.

With the popularity of unmanned aerial vehicle (UAV) technology, remote sensing target tracking in aerial videos from UAVs has drawn much attention. However, various problems occur, such as obvious target scale changes and frequent target shape updates in UAV aerial videos, and typical tracking algorithms have difficulties in solving these challenges. Therefore, we propose a remote sensing target tracking method for UAV videos based on a multiscale antideformation network (MSADN). This method uses the fully convolutional Siamese network (SiameseFC) as its basic architecture. First, in the target feature extraction stage, we use a receptive field block module to change the single convolutional layer of the network. And the multibranch structure can improve the algorithm’s adaptability to target scale changes. Then, in the tracking stage, to solve the problem of the lack of template updates, we integrate a template dynamic update module into the SiameseFC architecture. This module uses long short-term memory to generate control signals to dynamically update target shape information, and it will improve the algorithm’s ability to cope with frequent target shape updates. Finally, compared to state-of-the-art trackers, experimental results show that the MSADN algorithm can obtain better performance (75.9% Prec, 56.4% Succ) while ensuring higher efficiency (60 fps).

The objective of this work was to investigate the hypothesis that oceanic internal waves could modulate the production of atmospheric aerosols by modulating sea surface roughness. Airborne profiling lidar was used to measure the displacement of the subsurface scattering layer by internal waves, the distribution of surface roughness, and the aerosol concentration over the internal waves. The direct correlation between internal wave displacement and aerosol density was not statistically significant, primarily due to other processes causing mixing in the atmosphere. However, we found that the magnitude of the aerosol spatial power spectral density at the internal wave wavelength was significantly correlated with the internal wave magnitude. We developed a simple model to describe the interactions, which provided excellent agreement with the measurements for two out of three flights. The disagreement for the third flight is thought to be a result of the deeper thermocline on that day, which put the deeper parts of the internal waves out of the depth range covered by the lidar. The conclusion is that internal waves affect aerosol production over the ocean under certain environmental conditions.

Rapid and accurate crop information extraction is important for detailed agricultural management and efficient yield estimation. However, the natural environment of the Karst Plateau in southwest China is fragile, the ground surface is broken, and the weather is complex and cloudy, making it difficult to obtain high-quality crop samples for crop information extraction in this complex environment. We obtained images of Pitaya trees from plateau mountain environments using DJI Mavic 2 Pro UAV, constructed a small UAV close-range acquisition sample dataset, which included initial, supplementary, and augmented datasets, covering samples in complex natural scenes such as twining vines, weed and tree cover, blurred images, and shadows. We studied the influence of complex scenes on the extraction accuracy of Pitaya trees using the U-Net model to accurately delineate Pitaya trees in complex UAV images. The results show: (1) the U-Net model trained by the augmented dataset had the highest recognition precision of 99.20% for Pitaya trees, F1-score of 96.66%, and Kappa coefficient of 0.91. (2) The number of samples and the complexity of land types had strong impact on the recognition accuracy. From 200 to 21,593 samples, the accuracy of the recognition results, F1-score and Kappa coefficient increased by 17.47%, 17.95%, and 0.26%, respectively. Moreover, the missed detection rate significantly decreased (18.27% to 0.80%), the false alarm rate (5.36% to 1.04%). (3) When the sample types were increased from 1 to 10, the learning of sample features by the U-Net model, including shadows, blurred images, and twining vines, was strengthened. This enhanced the robustness and generalization ability of the model. The small sample dataset in this study meets the requirements of identifying and extracting information for Pitaya trees from the background of the rugged terrain and complex features in plateau and mountain areas.

Current estimates of stand volume for South Korean forests are mostly derived from expensive field data. Techniques that allow reducing the amount of ground data with reliable accuracy would decrease the cost and time. The fifth National Forest Inventory (NFI) has been conducted annually for all forest areas in South Korea from 2006 to 2010 and using these data we can make a model for estimating the stand volume of forests. The purpose of this study is to test deep learning whether it is available for measurement of stand volume with satellite imageries and geospatial information. The spatial distribution of the stand volume of South Korean forests was predicted with the convolutional neural networks (CNNs) algorithm. NFI data were randomly sampled for training from 90% to 10%, with 10% decrement, and the rest of the area was estimated using satellite imagery and geospatial information. Consequently, we found that the error rate of total stand volume was <5  %   when using over 17% of NFI data for training (R²  =  0.96). We identified that using CNNs model based on satellite imageries and geospatial information is considered to be suitable for estimating the national level of stand volume. This study is meaningful in that we (1) estimated the stand volume using a deep learning algorithm with high accuracy compare with previous studies, (2) identified the minimum training rate of the CNNs model to estimate the stand volume of South Korean forest, and (3) identified the effect of diameter class on error hotspots in stand volume estimates through clustering analysis.

We aim to classify acoustic events recorded by a fiber optic distributed acoustic sensor (DAS). We derived the information from probing the fiber with light pulses and analyzing the Rayleigh backscatter. Then, we processed this data by a pipeline of processing algorithms to form the input for our machine learning classification model. We put random matrix theory to the test to distinguish the acoustic event of interest from the noise. We conditioned the raw trace using moving average and wavelet-based filtering algorithms to improve the signal-to-noise ratio. For raw, low pass, and wavelet-based filtered signals that we inject into a convolutional neural network, we rely on the magnitude of their complex coefficients to categorize the nature of the event. We also investigate Mel-Frequency Cepstral coefficients specific to the event as an input for the classifier and compare their performance to other signal representations. We run the experiments on the CNN for two-class and three-class classification using datasets from a DAS that is deployed for perimeter security and pipeline monitoring. We obtained the best results when using the MFCCs paired with wavelet denoising, achieving accuracies of 96.4% for the “event” class and 99.7% for the “no event” class when it comes to the two-class process. The three-class process yielded optimal accuracies of 83.3%, 81.3%, and 96.7% for the “digging,” “walking,” and “excavation” classes, respectively. Finally, the training execution time is exceptionally long because the extensive dataset and the model’s architecture are complex. As a result, we make efficient use of the CPU and GPU to maximize our machine’s power using the Keras API’s sequence data generator. Compared with the serial implementation, we report an improvement of up to 4.87 times.

The existing point-scale or large-scale soil moisture content (SMC) monitoring techniques cannot entirely satisfy the requirements of in situ SMC monitoring for agricultural purposes. We proposed a framework for in situ mesoscale SMC monitoring based on global navigation satellite system interferometric reflectometry (GNSS-IR) and ensemble modeling technique. The framework consists of a data collecting and preparing system and an SMC retrieval system. The amplitude, phase, and detection depth, which are the GNSS-IR signal-to-noise ratio parameters, were positively and negatively correlated with the volumetric SMC data, respectively. In the SMC retrieval stage, five kinds of models including linear regression, multilinear regression, k-neighbor regressor, support vector regressor, and random forest (RF) was tested as the first-order model. Two kinds of swarm optimization algorithms including sparrow search algorithm (SSA) and particle swarm optimization (PSO) were examined for models’ hyper-parameter optimization. The results show that the RF performed best and had a coefficient of determination (R²) of 0.798, a root mean square error (RMSE) of 0.043  cm³  /  cm³, and a mean absolute error (MAE) of 0.034  cm³  /  cm³ for the validation set. Both the SSA and PSO are effective for models’ optimization. After input variable selection, the second-order ensemble RF model outperformed the first-order RF model and had an R² of 0.819, an RMSE of 0.040  cm³  /  cm³, and an MAE of 0.031  cm³  /  cm³ for the validation set. The proposed framework is potentially valuable for popularization because of its cost-effectiveness and high accuracy.

Aircraft detection in optical remote sensing images has important practical value and attracts great attention in recent years. Due to some clustered backgrounds and various deformations like variability in viewpoints, scales, and directions, it is still a big challenge. In recent years, deep learning, especially convolutional neural networks (CNNs), has achieved remarkable success not only in natural image object detection tasks but also in remote sensing image object detection tasks. However, due to the limitations of the actual application deployment environment only limited processing units and storage space can be used, and most deep learning-based aircraft detection algorithms cannot work well. To this end, a unique aircraft detection framework in optical remote sensing images is proposed, which is based on fast feature pyramids and a specialized lightweight CNN. The algorithm can be summarized into two steps. In the first step, a highly reliable region proposal generation algorithm is designed to predict region proposals. In this phase, a prior square sliding window method and an efficient linear search algorithm are proposed. In the next step, a lightweight CNN is designed to achieve fast and accurate aircraft classification. The proposed model only requires about 237 KB of parameter storage and can infer in a poor central processing unit platform. Comprehensive experiments on three publicly optical remote sensing aircraft datasets demonstrate the superiority and effectiveness of the proposed method. It is also shown that the proposed method is light and fast in comparison with some state-of-the-art methods.

The planetary boundary layer height (PBLH) is an essential parameter for weather forecasting and climate modeling. The primary methods for obtaining the PBLH include radiosonde measurements of atmospheric parameters and lidar measurements, which track aerosol layers in the lower atmosphere. Radiosondes provide the parameters to determine the PBLH but cannot monitor changes over a diurnal cycle. Lidar instruments can track the temporal variability of the PBLH and account for spatial variability when operated in a network configuration. The networkable micropulse DIAL (MPD) instruments for thermodynamic profiling are based on diode-laser technology that is eye-safe and cost-effective and has demonstrated long-term autonomous operation. We present a retrieval algorithm for determining the PBLH from the quantitative aerosol profiling capability of the high spectral resolution channel of the MPD. The PBLH is determined using a Haar wavelet transform (HWT) method that tracks aerosol layers in the lower atmosphere. The PBLH from the lidar is compared with the PBLH determined from potential temperature profiles from radiosondes. In many cases, good agreement among the PBLH retrievals was seen. However, the radiosonde retrieval often missed the lowest inversion layer when several layers were present, while the HWT could track the lowest layer.

Low elevation angles (between −3  deg and 3 deg) under bright temperatures (LEATBs) have not been efficiently employed in microwave remote sensing. LEATB data with horizontal polarization in the X-band and sea surface wind speed (SSWS) data, measured during an experiment in 2019, showed that when the SSWS is ≤4.9  m  /  s, the relationship between the brightness temperature ratio and SSWS is the opposite of that when the SSWS is over 4.9  m  /  s. Investigating the causes of this event, a LEATB-based simplified SSWS remote sensing method was developed. The results demonstrated that when the SSWS is below 4.9  m  /  s, the root mean squared error (RMSE) of the SSWS sensing is as low as 0.9544  m  /  s. However, when the SSWS exceeds 4.9  m  /  s, the corresponding RMSE is bound to 1.1349  m  /  s.

A timely and accurate understanding of the spatial distribution of tea plantations is beneficial for agricultural management and regional sustainable development. However, obtaining detailed distribution data on large-area tea plantations remains challenging owing to limitations in computational capabilities, training data, and workflow design. Utilizing the Google Earth Engine, which provides a catalog of multisource data in a cloud-based environment, we developed a methodology to generate a highly accurate tea plantation map, with a 10-m resolution, for Anhui Province, China, by integrating a random forest model with a progressive model. Our major contribution lies in this hybrid approach, which comprises two major components: (1) an optimal classification band combination derived from Sentinel-2 products and the digital elevation model filtered by the J-M distance model and (2) a progressive random forest method introduced for tea plantation classification. The experimental results show that our proposed workflow achieved an average classification accuracy of 89.27% for the entire Anhui Province. In addition, this approach is semiautomatic and can effectively reduce the labor required during the generation of training data compared with traditional classification approaches. These findings demonstrate the potential of integrating machine learning and progressive models to produce high-precision remote sensing classification maps.

Rapid urban growth and globalization affect land use in cities, and the need for automatic interpretation of remote sensing images is constantly increasing. Deep neural networks are becoming widespread in high-resolution aerial and satellite image sources in Earth observation missions. Various convolutional neural network (CNN) architectures have been implemented in building extraction, but it is still challenging to distinguish building class from other man-made classes in public datasets. Here, we present comparative research for automatic building extraction on different data sources using DeepLabV3+ architecture with ResNet-18, ResNet-50, Xception, and MobileNetv2 models. The CNNs are implemented on Inria Aerial Image Labeling, Massachusetts Buildings, and Wuhan University Building Extraction Datasets in terms of evaluation metrics and training and testing time consumption. Our implementation of the DeepLabV3 + ResNet-50 model performed F1-score of 97.44% in Massachusetts Building dataset and intersection over union as 80.75% in Inria dataset, higher than at least 3% than the previous studies.

We propose an innovative method for ship detection in the real marine environment. Different from other high-resolution optical images and synthetic aperture radar images, ship detection of coast defense radar is quite challenging due to the complex background, sea state, and low resolution. To this end, we build a real dataset and propose an innovative detection method based on you only look once (YOLO) V4. Specifically, first, the lightweight architecture MobileNetV3 is introduced as the backbone feature extractor to accelerate the detection speed by compressing the parameters. Second, for better detection of small-size ships, the squeeze-and-excitation module is used to apply the attention mechanism to the channel. Meanwhile, the scaled exponential linear unit non-liner activation function replaces the rectified linear unit activation function of the MobileNetV3 shallow layer, which optimizes the convergence effect of the model. Third, an adaptive anchor-selection algorithm for the detection of ships with various shapes is designed. Compared with other well-established models based on convolutional neural network (CNN), including single shot multi-box detector, Faster region based-CNN, and you only look once version 4 baseline for detecting ships, our improved method yields impressive results in our dataset. After extensive testing, the mean average precision of the proposed method can reach 97.43%, with the detection time per frame reaching 38 ms.

Land subsidence is generally caused by stress concentration of the overlying strata, which may reach the rock failure limit, causing surface collapse. The Liaohe Plain is an important resource producing area in China. Concentrated exploitation of underground resources leads to the generation of underground cavities, and surface subsidence gradually occurs as cavities increase. We used small baseline subsets synthetic aperture radar interferometry to monitor large-scale surface deformation on the Liaohe Plain. Field investigations found that the spatial distribution of land subsidence on the Liaohe Plain is highly correlated with coal mining areas. Taking the interferometric synthetic aperture radar deformation as the real constraint, a nonlinear Bayesian inversion algorithm was used to reconstruct underground reservoir parameters, such as the position and attitude of underground coal seams, based on a multi-source dislocation model. The R² between observed and modeled deformation in multiple coal mining areas was above 0.9. On this basis, relying on the Coulomb failure stress (CFS) criterion, the elastic model parameters and regional geological data were used to calculate the stress field caused by underground resource exploitation on the ground surface. The results show that the CFS fields of multiple groups of coal seams at close distances have mutual influence, and it is more concentrated at the perimeters than in the central portions of the subsidence area. The proposed dislocation model of closely spaced multiple coal seams improved the inversion accuracy of complex mining conditions, and would provide important information for risk assessment and prevention from the perspective of geodesy.

The automatic segmentation of buildings from satellite images is extremely challenging due to the complex shapes of buildings. Deep convolutional neural networks (DCNNs) have recently enabled accurate pixel-level labeling tasks, which ensures precise boundaries. However, the large number of pixel-level labels required for DCNN-based semantic segmentation of buildings represents a considerable obstacle to the process. We propose a framework that relies on deep seeds and optimal segmentation to extract buildings from very high resolution imagery to solve the aforementioned issue. Input images are cropped rather than resized to a low-resolution image and passed directly to DCNN for segmentation. An image resize transformation will lead to a significant loss of image detail, especially for images with high resolution. For both image patches and resized images, a classification network is utilized to locate deep seeds within building and nonbuilding classes. Then, a boundary map can be predicted by passing the resized image through a convolutional oriented boundary network. A hierarchical segmentation tree is built to determine the optimal segmentation by seeking the optimal tree cut. The final segmentation is achieved by developing a graphic model representing a set of nodes that distribute information from deep seeds to unmarked regions. According to the experiments conducted on the ISPRS two-dimensional semantic labeling contest (Potsdam) and the WHU building datasets, the proposed framework is able to significantly enhance the building segmentation accuracy. We have achieved such improvements through a cost-effective computing method without training a segmentation network.

The Tamera open pit mine is located in the Nefza mining district (north-west of Tunisia) where the reddish soil reveals its strong potential for heavy Fe mineralization, along with other metals (Pb, Zn, and Mn). This study explored Iron oxide mineralization in a particular nappe zone using multiscale field visible and near infrared/short waves infrared and EO-1 Hyperion hyperspectral data. Thus, spectral features fittings were conducted on Hyperion data for mineral mapping using scene-extracted endmembers. A first minimum noise fraction algorithm was used to reduce the hyperspectral data’s dimensionality, followed by the Pixel Purity Index and n-dimensional visualization to extract the pure pixels (endmembers) from the cluster of pixels in the data cube. Furthermore, 56 soil samples were collected from the test site and spectrally measured using Analytical Spectral Devices FieldSpecHiRes spectrometer, which covers the spectral range between 350 and 2500 nm. All samples were analyzed in the laboratory with x-ray diffraction. Endmember extracted spectra were compared with field measurements before EO-1 Hyperion processing. Mapping results revealed the occurrence of a mineralogy composition consisting particularly of iron oxide minerals (hematite and goethite) and bearing minerals, such as clays (kaolinite), within all samples.

There are often significant intensity variations between multispectral images, making automatic registration tasks difficult. Traditional feature matching methods, such as the scale-invariant feature transform (SIFT), are often sensitive to nonlinear variations of intensity between multispectral images. To solve this problem, an improved SIFT algorithm is introduced. First, the contrast limited adaptive histogram equalization algorithm is introduced in the feature extraction stage to improve the feature point extraction results. Then, the Sobel operator is used to enhance the main direction consistency between homologous feature point pairs. The experimental results suggest that the method can obtain reliable registration results on unmanned aerial vehicle multispectral images.

The detection and recognition of aircraft on images is an important task in the field of optical remote sensing image interpretation. The challenges include the detection of arbitrary orientation targets and the fine-grained recognition of aircraft type. However, most related studies only pay attention to one of these because of the limitation of datasets, and an end-to-end network often cannot simultaneously solve these challenges well. Therefore, we propose a cascade framework for arbitrary-oriented and multitype aircraft detection in remote sensing images based on the convolutional neural network. A rotation-based detector that uses oriented boxes is designed to determine the position and direction of aircraft on remote sensing images. A fine-grained recognition subnetwork that adopts the ensemble learning and the Fisher discriminant regularization is designed to recognize the type of aircraft on images. The subnetwork is cascaded behind the rotation-based detector to achieve more accurate recognition. The mean average precision of this method reaches 83.7% on a challenging dataset, which is 2.4% to 8.2% higher than other advanced methods. Experiments demonstrated that the proposed method can greatly improve the accuracy of aircraft detection and recognition and that the method has good generalization performance.

Glaciers in the Himalayan mountain ranges contain huge amounts of water and are essential for the water security of their downstream basins. Due to the spatiotemporal heterogeneity and uncertainty of the glacier mass change (GMC) in the Himalayas, more refined and deeper investigations on the GMC and its driving forces are needed to complement the current research. We present the mass balance of Himalayan glaciers during 2000 to 2013 using the latest global digital elevation models; this method focuses on establishing a high-coverage result of the GMC and resolving the problems of timestamps and penetration uncertainties. The GMC of the entire Himalayas is −0.32  ±  0.15  m w.e.yr^−  1; most of the lost mass drains into the Indus and Ganges river basins. Glacier mass loss gradually decreases from east to west, with the East and Central Himalayas exhibiting the highest heterogeneity in the vertical and horizontal directions, respectively. We propose two multivariate regression models with 12 potential driving factors that explain 63% to 89% of the GMC across different regions of the Himalayas. Precipitation significantly influences changes in glacier mass throughout the Himalayas, whereas debris coverage and glacier size have insignificant effects on changing the glacier mass. Our analysis reveals that seasonal climate-driven changes in the glacier mass of the Himalayas may threaten the downstream water supply.

Mercury (Hg) is a heavy metal pollutant characterized by global migration, high bioaccumulation, and high biological toxicity. Soil Hg pollution seriously harms soil environmental quality and human health. The higher the content of Hg in the soil, the more serious the harm. Therefore, the acquisition of soil Hg content information is crucial for soil monitoring and restoration, and human health. The purpose of this study is to establish the best hyperspectral prediction model of soil Hg content in the study area. A total of 90 soil samples were collected from the mining areas and suburbs, and the Hg content was analyzed. Several preprocessing methods were applied to the hyperspectral data. Partial least-squares regression (PLSR), orthogonalized PLSR, random forest, and back propagation neural network (BPNN) were used to establish the relationship between soil hyperspectral data and soil Hg content to predict the soil Hg content. Specifically, the transferability of the best prediction model in yellow-brown and cinnamon soil data was discussed. The results showed that BPNN + standard normal variable + full-band (4 nm) is the optimum method to predict soil Hg content (Rtrain2=0.94, Rtest2=0.85). The transferability of the model is the best in yellow-brown soil. This research provides technical support for the rapid monitoring of soil Hg pollution information in central and eastern Shaanxi-Province, and the results enrich prediction models of soil Hg content in different soil types.

One of the greatest challenges in many machine learning applications is to build robust classification models that reduce calibration efforts, ensure model transferability to unseen data distributions, and handle high semantic resolution datasets with a relatively large number of target classes. We present an operational domain adaptation framework for crop type verification from remote sensing data for integrated administration and control and farm management systems. It is part of the ag|knowledge crop monitoring platform and demonstrates a more robust spatiotemporal model transferability than conventional supervised crop type classification approaches, which use multitemporal and multispectral features to classify the specific characteristics of crop phenology patterns. The proposed method is based on a machine learning model that has been trained on particular similarity metrics between all target crop types. The metrics are derived by quantifying the similarity of vegetation index time series between the observed parcel and the aggregated time series of reference objects for each parcel and crop type. This method achieves higher classification results than a model trained on the pure remote sensing time series data, as the algorithm does not learn the temporal vegetation index patterns, but instead, the adversarial characteristics and the differences between each crop type. Moreover, it also reduces dimensionality as the time series are summarized through the respective similarity metrics. We compare both approaches through different classification algorithms. Overall, the achieved classification accuracy for more than 67 crop types scores more than 80%.

The Moderate Resolution Imaging Spectroradiometer (MODIS) on board the Terra and Aqua platforms has 20 reflective solar bands (0.4 to 2.1  μm), for which the on-orbit detector gain changes are tracked using the on-board solar diffuser (SD) and SD stability monitor. SD and lunar observation data, supplemented by Earth-view (EV) observations for some bands, are used to characterize the scan-mirror response versus scan angle (RVS). The short-wave infrared (SWIR) bands (1.2 to 2.1  μm) have issues, such as electronic crosstalk and an out-of-band optical leak from the thermal bands, which complicate the use of the Moon for reliably characterizing the on-orbit RVS. Before Collection 7 (C7), the pre-launch RVS was deemed sufficient for both Aqua and Terra MODIS SWIR bands throughout the missions. However, recent studies have found that the RVS behavior does change with time in some SWIR bands, up to 2% and 1% for Terra MODIS bands 5 and 26, respectively. We assess the residual RVS effects present in the current MODIS calibration at desert sites and deep convective clouds (DCC). Due to the large uncertainties caused by the water vapor absorption effects from desert sites, especially for band 26 (1.375  μm), EV observations from DCCs are chosen to characterize the on-orbit RVS behavior for the SWIR bands. A time-dependent RVS characterization method for the SWIR bands is described and assessed by comparing the results with Collection 6.1 (C6.1) Level 1B (L1B) reflectance products. This new algorithm will be used in the upcoming C7 L1B reprocessing for Terra MODIS, providing more stable and accurate SWIR band reflectance data over time and across the scan range compared with the C6.1 results. For Aqua MODIS SWIR bands, performance evaluations with desert and DCC data indicate stable performance, so the pre-launch RVS will continue to be used in C7.

We present a one-bit quantified phase-based algorithm combined with the noiseless correlation transformation (PANCT) to estimate the three-dimensional (3D) parameters (i.e., azimuth angle, elevation angle, and range) of wideband near-field source via uniform circular array. We first propose a noiseless correlation transformation scheme without direction of arrival pre-estimation to obtain the correlation matrix of focusing frequency bin. Then, we exploit the phase differences to construct indefinite matrix and implement the least-square method to derive 3D parameters. Considering that the complexity scenarios include phase ambiguity and one-bit quantization, we further propose the corresponding unambiguous PANCT algorithm and one-bit quantified PANCT algorithm, respectively. Simulation results demonstrate that the proposed algorithms provide satisfactory estimation performance with low computational complexity.

Water conveyance tunnels perform an important role in water-diversion projects. With a water conveyance tunnel transferring water, the conflict between urban crowded ecology and agricultural water can be relieved. Research shows that there is a significant need for a more objective and accurate inspection of water conveyance tunnels during operation. Hence, we investigate the use of a single-beam scanning sonar to detect a water conveyance tunnel and generate a three-dimensional point cloud model, which can be used to make objective predictions for surface defects. First, a cobweb-like grid matrix was proposed to grid discrete points to establish the topological relationship of the point cloud. Second, a denoising method fusing acoustic properties and intensity was used to filter the noise in the acoustic point cloud. Third, an optimal echo extraction technology based on a second-order differential model was proposed to obtain a high-precision point cloud of the tunnel. Experiments show that the proposed denoising method and optimal echo extraction method are superior to traditional algorithms commonly used in ground point clouds, with good results achieved at real test sites.

Object detection is a technique used to localize and classify objects in an image or a video sequence. It is an emerging topic of research in the field of computer vision. However, detections in a video are affected by sensor-specific challenges. A convolutional neural network-based You Only Look Once, version 3 (YOLOv3) object detection algorithm was used to get optimized computation time and accuracy. In the proposed methodology, the YOLOv3 architecture extracted significant features from both the visible and thermal imaging domains, and an adaptive fusion of both domains was performed to determine the dominant imaging domain and provide robust detections. The resulting YOLOv3 detections included the bounding box coordinates, confidence score, and class from each imaging domain, which were fused implicitly into a single plane. The sensor domain having the maximum number of object detections was chosen as the reference to be compared with the other domain for the adaptive fusion process. After fusion, the algorithm removed redundancy using adaptive intersection over union thresholding. The mean average precision result obtained from the fusion algorithm was 44.25%. A comparative study was also carried out between pre-trained common objects in context weights and custom CAMEL dataset weights; it showed the significance of using adaptive fusion in challenging situations such as nighttime, shadow, varying illumination, moving camera, and crowd.

The generalized bilinear model (GBM) has been one of the most representative models for nonlinear unmixing of hyperspectral images (HSI), which can consider the second-order scattering of photons. Recently, robust GBM-low-rank representation (RGBM-LRR) for nonlinear unmixing of HSI has been introduced to capture the spatial correlation of HSI using LRR with nuclear norm minimization (NNM). However, NNM is used to approximate the matrix rank by shrinkage all singular values equally. The singular values have distinct physical significance in many real applications, and NNM may not be able to estimate the matrix rank accurately. To overcome the above issue, a robust GBM with a weighted low-rank representation (RGBM-WLRR) approach is proposed using weighted nuclear norm minimization, which mitigates the penalty on larger singular values by assigning a small weight, so the corresponding shrinkage is small and also takes serious shrinkage on small singular values by assigning larger weights to them. The proposed model is solved using an iterative alternating direction method of the multipliers. A series of experiments with real datasets and a simulated HSI with varying rank and signal-to-noise ratio reveals that RGBM-WLRR performs significantly better than the state-of-the-art algorithms in terms of signal-to-reconstruction error, root-mean-square error, and spectral angle distance.

Image registration plays an important role in surface detection, aerial image classification, and satellite image fusion. However, the existing image registration frameworks pay less attention to local feature information. Additionally, such frameworks have insufficient extraction of important channel and spatial information and ambiguous matching. To address these issues, we propose an end-to-end remote sensing image registration framework based on the improved attention neighborhood consistency network. Our network improves the robustness of the rotation distortion region by employing an asymmetric convolution block in the feature extraction stage, which is augmented by two one-dimensional convolution kernels with square convolution kernels. Moreover, we use a 1  ×  1 convolution to adjust the residual network channel to retain global information, realize feature reuse, and obtain multiscale image information. Next, the attention mechanism is improved by utilizing skip connections to resolve the problems of high resolution and image heterogeneity issues in remote sensing images. This allows the framework to quickly focus on important regional features and improve the network’s ability to distinguish and localize image features. In the feature matching stage, a bidirectional cross-correlation algorithm is used to establish the correspondence between the two pictures and match the corresponding points are matched. At the same time, the multilayer perceptron (MLP) structure is introduced into the neighborhood consensus network to eliminate the noise generated by mismatching. Further, it establishes a strong matching locality, prompts information aggregation at different positions and different channels, reduces registration deviation, and enhances matching robustness. The experimental results show that, overall, the algorithm outperforms comparable algorithms, extracting more accurate feature points and efficiently reducing the mismatch rate. The algorithm put forward effectively functions in a variety of situations, exhibiting high efficiency and high accuracy.

For passive bistatic inverse synthetic aperture radar (PB-ISAR) imaging of maneuvering targets, the complex motion of the target causes non-uniform time-varying bistatic angle and non-uniform rotation of the target, which leads to the severe bistatic distortion and defocusing in the generated ISAR image. To obtain a focused, well-scaled PB-ISAR image that can show the accurate shape and size of the target, the essential issues, including bistatic distortion, image defocus, and range and cross-range scaling (RCRS), must be addressed. An innovative PB-ISAR imaging and cross-range scaling method is proposed for obtaining ISAR image of the maneuvering targets under low SNR condition. First, we estimate the multi-dimensional parameters to approximate the non-uniform time-varying bistatic angle and non-uniform rotation of the target. Then, based on the estimated parameters, the bistatic distortion correction, defocus terms elimination, and RCRS can be performed subsequently. Finally, a focused and well-scaled PB-ISAR image of the target can be obtained. The proposed method can work under a low signal-to-noise ratio. The simulation results validate the effectiveness of the proposed method.

Vehicle detection by satellite videos is important for urban intelligent traffic construction. A satellite videos vehicle detection method is proposed to solve the problems of the surrounding background and the vehicles having a similar color, and the low number of pixels occupied by vehicles. First, we design a weighted normalization sparse autoencoder network, highlighting the effective features and strengthening the expression of the vehicle features by introducing weighted normalization sparse autoencoders into the backbone. Second, we design a method for the early fusion of spatial information transmission, realizing the enhancement of spatial information in the deep feature map through atrous convolution. Finally, we propose K-means++ based on adjusted cosine similarity, stretching the anchor box size by introducing the adjusted cosine similarity calculation. The experimental results show that the average accuracy of this method in satellite videos vehicle detection reached 75.7%, and the recall rate reached 89.6%. This method can effectively detect vehicles in satellite videos, and the effect is significantly improved compared with other methods.

The sizes of ships vary by type in synthetic aperture radar (SAR) images, and this multiscale problem degrades the accuracy of ship detection. We present a multiscale ship detection method based on cascaded dense weighted networks in SAR images. Dense weighted fusion is performed in the feature extraction module and feature fusion module by cascading, i.e., feature weighting is achieved by embedding a convolutional block attention module after feature layers at different scales, followed by dense connection to fuse features at different scales. This makes it possible to refine the cascaded feature mapping, enhance the information transfer of shallow and deep features, and build a multiscale feature representation model. Experiments on the extended high-resolution SAR images dataset show that the accuracy of this method is 74.87% in inshore scenes, and 97.39% in offshore scenes, which exceeds that of other ship detection methods. In addition, in two SAR images from TerraSAR-X, the detection accuracy of multiscale ships reaches 86.63%, which indicates the method’s better robustness.

Multispectral imaging can offer many benefits in cost, complexity, resolution, size, weight, and power, relative to hyperspectral imaging. When designing a multispectral system, spectral bandpasses can be selected using optimization algorithms configured to maximally separate target detection scores between target and background regions. A hyperspectral image (HSI) can serve as the source of data from which band groupings can be tested for optimality. The output of an adaptive cosine estimator target detection algorithm is used in an objective function. Three optimization algorithms are compared: particle swarm, dual annealing, and differential evolution. A global optimum is also found using a brute force approach on the Livermore Computing Syrah supercomputer. Three materials are investigated: calcite, gypsum, and limestone. This is done for 3-, 4-, and 5-band systems. The data originate from a longwave infrared HSI of a material display board. The optimization algorithms were run 30 times for every scenario. Performance statistics (maximum, minimum, mean, standard deviation, and median) based on the separation values are given. Additional characterization was performed using receiver operator characteristic (ROC) curves and the area under the ROC curve. While good performance was obtained for the three optimization algorithms, the dual annealing algorithm produced the highest and most consistent detection separation scores on average.

Convincing depth estimation from monocular aerial images is a fundamental task in environment perception. We propose an efficient multilevel feature fusion network (MLFFNet) for estimating depths from aerial images. Specifically, the proposed MLFFNet consists of a multi-level attention pyramid (MAP) module and an adaptive feature fusion (AFF) module. The MAP module extracts the low-level and high-level useful information from the perspective of nonlocal spatial attention and channel attention, while the AFF module adaptively integrates the extracted multilevel features to enhance the estimated effect. Moreover, since images taken by drones have a large depth range, we designed a loss function suitable for aerial images. The evaluation experiments are performed on the MidAir dataset. Experimental results denote that our MLFFNet outperforms other depth estimation methods in predicting the depths. We also test several images from real-life scenarios, and our method can obtain reasonable outputs.

Hyperspectral image (HSI) classification methods based on three-dimensional convolutional neural network (3DCNN) have problems of overfitting the in-sample training process and difficulty in highlighting the role of discriminant features, which reduce the classification accuracy. To solve the above problems, an HSI classification method based on M-3DCNN-Attention is proposed. First, the Mixup algorithm is used to construct HSI virtual samples to expand the original data set. The sample size of the expanded data set is twice that of the original data set, which greatly alleviates the overfitting phenomenon caused by the small sample of HSI. Second, the structure of 3DCNN is improved. A convolutional block attention module (CBAM) is added between each 3D convolutional layer and ReLU layer, and a total of three CBAMs are used so as to highlight the discriminant features in spectral and spatial dimensions of HSI and suppress the nondiscriminant features. Finally, the spectral–spatial features are transferred to the Softmax classifier to obtain the final classification results. The comparative experiments are conducted on three hyperspectral data sets (Indian Pines, University of PaviaU, and Salinas), and the overall accuracy of M-3DCNN-Attention is 99.90%, 99.93%, and 99.36%, respectively, which is better than the comparative methods.

To realize the digital elevation inversion of the interferometric imaging radar altimeter (InIRA), an interference complex images registration algorithm combining enhanced scale-invariant feature transform (SIFT) characteristics with correlation coefficient is proposed. First, the locally tuned nonlinear method is used to enhance the image features. Then, SIFT algorithm is used to extract the matched feature points that are used as control points after screening. Based on these control points, the affine transformation is applied to calculate the coarse matching relation. Second, multiple control points are chosen uniformly. The local accurate offsets are determined by interpolating and calculating the maximum of correlation coefficients. The least-squares method is used to fit the difference between the two images. Third, the two images are matched by interpolating and resampling the one to be registered. Finally, the simulated InIRA sea surface images and the Sentinel-1A images of the Mount Hua area are employed to experiment. The results show that the proposed algorithm combines the advantages of SIFT algorithm and correlation coefficient algorithm. It is robust and its registration accuracy is better than the particle swarm optimization sample consensus algorithm, unsupervised deep-learning algorithm, SIFT algorithm, and correlation coefficient algorithm.

Aiming at the problem of object blur in optical remote sensing images caused by atmospheric disturbance and cloud disturbance, a remote sensing image super-resolution reconstruction network, namely feature-enhanced reconstruction generative adversarial network (FR-GAN), based on a generative confrontation structure is proposed. The network adopts a self-supervised learning mode to achieve high-quality image reconstruction by learning the characteristics of the reconstructed image. First, to simulate the blurring effect of remote sensing images affected by interference factors, Gaussian noise is introduced into the image degradation model of self-supervised learning. Second, a background information enhancement module is designed in the FR-GAN image generator, which solves the problem of blurred objects in the reconstructed image by extracting the background contour information of the image and enhancing its expressive ability. Finally, the relative loss and perceptual loss are used to construct the loss function of FR-GAN. On the one hand, it improves the stability of the training process, and on the other hand, it improves the quality of image detail texture reconstruction. The experimental results show that, compared with the existing super-resolution reconstruction networks, the image reconstructed by FR-GAN has a clear texture, rich details, and better subjective visual effect.

Recognizing land type using remote sensing images is vital for land resource management and monitoring. To improve the accuracy of land scene recognition from remote sensing images captured by sensors, an innovative classification method based on an improved capsule network was proposed. First, a circular local binary pattern algorithm extracted texture features from remote sensing images. The textured images were then fused with the original remote sensing images, and a 1-by-1 convolutional layer combined the fused feature graphs linearly. Finally, an improved capsule network performed remote sensing image scene classification. The experimental results show that the average accuracy rate of the proposed method improved by more than 6% compared with that of the typical convolutional neural network method, and the proposed approach outperformed other comparable methods in terms of accuracy. The proposed method not only provides a fast convergence speed but also improves the scene classification accuracy for remote sensing images, especially for images with rich texture information. This is helpful for land remote sensing image scene classification in various applications.

To improve the target recognition and classification performance of integrated radar and communications system in signal-dependent interference, we propose an innovative robust adaptive orthogonal frequency division multiplexing integrated radar and communications waveform design taking into account the clutter. We set the uncertainty class for clutter and other prior information to solve the problem of system performance degradation caused by inaccurate information in actual environments. Then, we formulate the optimization function to maximize the normalized weighted sum of mutual information and data information rate subject to the constraint of the total transmitted energy. To solve the optimization problem, we propose a maximum marginal allocation algorithm that transforms the original problem into a multi-stage energy distribution problem, and performs optimal energy distribution in each stage according to the optimization criteria. Moreover, by considering a more practical scenario of multiple downlink communication users, a joint scheme for channel allocation and waveform design is presented to ensure the fairness of multiple communication users. Simulation results show the effectiveness and feasibility of the designed waveform and channel allocation method.

The range resolution of inverse synthetic aperture radar (ISAR) imaging can be improved by directly increasing the bandwidth of the transmitted signal. However, it complicates the design of radar system hardware and increases the manufacturing cost. Aiming at solving the abovementioned problems, a multi-band ISAR fusion imaging method based on the multiple measurement vectors (MMV) model is proposed to improve the range resolution. First, a multi-band ISAR fusion imaging model based on the compressed sensing theory is established. Second, to improve the computational efficiency, a MMV accelerated improved linearized Bregman algorithm is proposed to solve the model. Nesterov’s acceleration gradient method and the condition number optimization of the sensing matrix are combined to further improve the iterative convergence speed. Finally, experimental results based on the simulation data and measured data verify the effectiveness and superiority of the proposed algorithm, which can achieve multi-band ISAR fusion imaging with higher imaging efficiency and better image quality.

Hyperspectral image (HSI) classification is a focus area in remote sensing research, wherein redundant spectral information poses a significant challenge and deep-learning-based classifiers have achieved better performance than traditional methods have. Training a deep-learning-based classifier requires numerous labeled samples. However, collecting such a substantial amount of labeled hyperspectral data is difficult. Semisupervised classification of HSIs has thus received increasing attention, where semisupervised learning classifiers function based on labeled and unlabeled data. A new training method for semisupervised HSI classification is proposed. Specifically, consistency regularization and pseudolabeling are combined as a semisupervised training framework, without the introduction of a complex mechanism. Our proposed algorithm can work without the need to change the conventional convolutional neural network model architecture. Unlike previous deep-learning-based methods, our approach does not require data reconstruction to obtain unsupervised loss. This means that our model can be much less computationally intensive. From the results of experiments on three public hyperspectral datasets, our proposed method outperforms several state-of-the-art methods.

To explore potential applications of GaoFen-3 (GF-3) fully polarimetric synthetic aperture radar (SAR) data in ground-feature classification, we selected Jiangxiang Town, Nanchang County as the research area and used GF-3 quad-polarization strip I (QPSI) data for classification. Pauli decomposition, Krogager decomposition, and the H  /  A  /  α decomposition model were used to decompose the GF-3 full-polarization data; combined with the backscattering coefficient values of GF-3’s HH, HV, VH, and VV polarization data, feature optimization was performed, and the optimal ground-object features for feature classification were obtained. The random forest (RF) method was used to classify the optimal features, and the results were then compared and verified. The results demonstrate that the component after polarization decomposition of H  /  A  /  α is more important in classification than the original backscattering coefficient values of HH, HV, VH, and VV. GF-3 single-phase SAR data were used for ground-feature classification in a crop-intensive area of South China, and the optimal feature optimization results included H  /  A  /  α decomposition and synthesis components of the alpha, anisotropy, red, entropy, blue, green, and polarization backscattering coefficients of VV. The RF method’s classification accuracy was higher than that of the maximum-likelihood, support-vector-machine, and neural-network methods. Using this approach, GF-3 data classification’s overall accuracy was up to 78.17%, and the Kappa coefficient was 0.738. This study has clear practical significance and reference value for the selection of classification methods for ground features based on GF- 3 data in cloudy and rainy areas.

We propose a moving target detection (MTD) and parameter estimation algorithm based on the range frequency reversing transform (RFRT) and second-order Wigner–Ville distribution (SoWVD) to solve the problem of range migration and Doppler migration of air-moving targets during long-term coherent integration. The algorithm first performs the RFRT before multiplying it with the original data to remove range migration. Next, the fast Fourier transform is applied to realize the SoWVD algorithm to quickly estimate the target’s acceleration parameters. Finally, after compensating for the acceleration parameters, the scaled inverse Fourier transform (SCIFT) is employed with the target’s initial range and velocity. In the simulation experiment, compared with five representative algorithms—MTD, SCIFT, adjacent cross-correlation function and Lv’s distribution, three-dimensional scaled transform, and Radon–Fourier transform and keystone transform—the RFRT-SoWVD algorithm achieves a good balance between computational cost and detection probability. Moreover, we utilize real measured radar data of an unmanned aerial vehicle target to verify the RFRT-SoWVD algorithm. The results show that the proposed algorithm can achieve excellent detection and parameter estimation performance.

Building extraction from high-resolution remote sensing images is the foundation of many fields, but the existing algorithms cannot extract building details from remote sensing images well because of complex background conditions and the occlusion of interfering objects. To solve the problems of low accuracy and incomplete boundary of traditional building extraction methods, ProCBAM, a parallel attention mechanism based on U-Net network, is adopted and added to the feature transmission step of U-Net. Through the representation of spatial dimension image feature map and channel dimension optimized feature map, we can learn more detailed image information and reduce the error of image recognition. Experiments are carried out on the Massachusetts building dataset, Wuhan University dataset, and IND.v2 dataset, and the experimental results show the effectiveness of this method in building extraction from remote sensing images.

In maritime management, satellite videos with a wide imaging range can provide beneficial data supplements for ground monitoring systems and improve the comprehensive management level. But there are few ships detailed features and a large amount of similar shape information in ship tracking by satellite videos, which lead to tracking drift even false tracking. A ship tracking method is proposed to resisting the interference information of ship-similar shapes. First, the mutual convolution Siamese network is used to calculate the similarity between the object template and the search area to improve the significance of the ship in the feature map. Second, the region proposal network with hierarchical double regression is proposed. Double regression of candidate boxes according to intersection over union (IOU) thresholds can reduce the influence of the non-rigid motion of the water surface on tracking and improve the IOU between candidate boxes and object ships. Finally, the object discrimination model is designed to enhance the screening ability of candidate boxes using deep and shallow classification head fusion and online training. To verify the effectiveness of the proposed method, tests were conducted on Jilin-1 satellite videos and public remote sensing data sets. The results show that, compared with the existing methods, the proposed method can effectively reduce the interference of similar shape information to ship detail features, and the tracking accuracy and success rate are 83.4% and 58.9%, respectively.

As a part of NASA’s Earth Observing System (EOS), the Terra spacecraft was launched on December 18, 1999, with the goal of understanding the changes of the Earth, by examining the Earth’s hydrological, geophysical, and climatic processes. The Moderate-resolution Imaging Spectroradiometer (MODIS) and Multi-angle Imaging Spectroradiometer (MISR) instruments on the Terra platform, combined with their continuous operation, broad spectral coverage, and different spatial resolutions, have played an important role to achieve the goals of the EOS. Over two decades of successful operations, these multispectral imaging instruments have benefited a variety of scientific applications. Being on the same platform, the two sensors complement each other in terms of spatial coverage (and target viewing geometry) and facilitate synergistic applications using multispectral data. A consistent radiometric calibration between these sensors is a prerequisite for creating high-quality science products from their observations. Both instruments underwent intensive prelaunch characterization and their on-orbit calibrations are monitored using their onboard calibrators. We perform a calibration inter-comparison of the spectrally matching bands of the two instruments using vicarious techniques. These techniques include multiyear simultaneous views of the North African desert, North Atlantic Ocean, and Dome Concordia; therefore, covering different reflectance regimes. Moreover, the near-simultaneous top-of-atmosphere (TOA) reflectance measurements from Railroad Valley (RRV), as provided by the radiometric calibration network (RadCalNet) (converted to TOA), are also included, which is used as a calibration reference to compare the on-orbit observations between MODIS and MISR. Simultaneous overpasses from desert, ocean, Dome C, and RadCalNet over RRV reveal that the agreement between the four spectrally matching bands is within 3% for the time-period between 2014 and 2020. Furthermore, some long-term drifts are observed in the TOA reflectance time-series from MISR for the red and NIR bands that are expected to be corrected in a future calibration reprocess.

Benefiting from the development of hyperspectral imaging technology, hyperspectral image (HSI) classification has become a significant research direction in remote sensing image analysis. However, labeling HSI requires sufficient domain knowledge and consumes a lot of human and material resources. In addition, HSI contains rich spectral and spatial features, and how to fully extract these two joint features remains a problem worth exploring. A semi-supervised spatial–spectral method based on three-dimensional (3D) Gabor and co-selection self-training (ST) is proposed for HSI classification. First, to fully extract the spatial–spectral features, the 3D Gabor filter is used for the raw HSI data cube to generate a Gabor feature data cube with spatial–spectral joint features. Second, the Gabor feature data cube is fed into the ST process to obtain sufficient labeled samples. In the ST process, we combine the classification results and clustering results to construct the candidate sample set. Finally, a co-selection strategy is proposed to select highly confident pseudo labels from the candidate sample set, and these selected pseudo labels are added to the labeled samples set by an iteration process. The experimental results obtained with two HSI datasets with different scenes (i.e., agricultural scenes and urban scenes) show that the proposed method has a stable classification performance in a few initial labeled samples.

A method to calculate the altitude-dependent, vertical, cloud-free line-of-sight (CFLOS) using the fractional cloud cover from the European Centre for Medium Range Weather Forecasting Re-Analysis Interim (ERA-I) dataset has been developed. This method enables users of airborne and satellite collections of optical ground data to understand the statistical coverage limitations of these collection systems by informing them of global probabilities of CFLOS versus altitude as well as time of year. This method is accurate for regions between ±60  deg of latitude; it should not be applied to polar regions due to limitations in the underlying ERA-I data. Our CFLOS calculations have been compared to the results of CloudSAT/Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation analysis and with Moderate Resolution Imaging Spectroradiometer (MODIS) total cloud cover data. It is shown that the ERA-I reports on average less cloud cover by about 7.5% (absolute) for regions within   ±  60 deg of the equator relative to MODIS cloud cover retrievals. Our CFLOS calculation leverages the resolution and diversity of ERA-I that enables spatial coverage as well as frequency of occurrence CFLOS calculations for nearly all non-polar regions on earth.