Search

The difficulty of very high resolution image change detection lies in the low interclass separability between the changed class and the unchanged class. According to experiments, we found that this separability can be improved by mining the relationship contained in the training samples. Based on this observation, a supervised change detection approach is proposed in this paper based on relationship learning. The proposed approach begins with enriching the training samples based on their neighborhood relationship and label coherence; this relationship is then learned simultaneously with the classifier, and, finally, the latter classification performance benefits from the learned relationship. Experiments demonstrate the effectiveness of the proposed approach.

Stereo matching is a fundamental task for 3D scene reconstruction. Recently, deep learning based methods have proven effective on some benchmark datasets, such as KITTI and Scene Flow. UAVs (Unmanned Aerial Vehicles) are commonly utilized for surface observation, and their captured images are frequently used for detailed 3D reconstruction due to high resolution and low-altitude acquisition. At present, the mainstream supervised learning network requires a significant amount of training data with ground-truth labels to learn model parameters. However, due to the scarcity of UAV stereo matching datasets, the learning-based network cannot be applied to UAV images. To facilitate further research, this paper proposes a novel pipeline to generate accurate and dense disparity maps using detailed meshes reconstructed by UAV images and LiDAR point clouds. Through the proposed pipeline, this paper constructs a multi-resolution UAV scenario dataset, called UAVStereo, with over 34k stereo image pairs covering 3 typical scenes. As far as we know, UAVStereo is the first stereo matching dataset of UAV low-altitude scenarios. The dataset includes synthetic and real stereo pairs to enable generalization from the synthetic domain to the real domain. Furthermore, our UAVStereo dataset provides multi-resolution and multi-scene images pairs to accommodate a variety of sensors and environments. In this paper, we evaluate traditional and state-of-the-art deep learning methods, highlighting their limitations in addressing challenges in UAV scenarios and offering suggestions for future research. The dataset is available at https://github.com/rebecca0011/UAVStereo.git

This special issue of the IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing (JSTARS) contains 11 papers both from the extended outcomes of the Multitemp 2019 presented papers, and from the submissions by following a general Call-for-Papers of this special issue. These papers focus on the interesting and relevant topics in the multitemporal data analysis.

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

Several machine-learning algorithms have been proposed for remote sensing image classification during the past two decades. Among these machine learning algorithms, Random Forest (RF) and Support Vector Machines (SVM) have drawn attention to image classification in several remote sensing applications. This article reviews RF and SVM concepts relevant to remote sensing image classification and applies a meta-analysis of 251 peer-reviewed journal papers. A database with more than 40 quantitative and qualitative fields was constructed from these reviewed papers. The meta-analysis mainly focuses on 1) the analysis regarding the general characteristics of the studies, such as geographical distribution, frequency of the papers considering time, journals, application domains, and remote sensing software packages used in the case studies, and 2) a comparative analysis regarding the performances of RF and SVM classification against various parameters, such as data type, RS applications, spatial resolution, and the number of extracted features in the feature engineering step. The challenges, recommendations, and potential directions for future research are also discussed in detail. Moreover, a summary of the results is provided to aid researchers to customize their efforts in order to achieve the most accurate results based on their thematic applications.

The increment in the data rate of the new-generation hyperspectral sensors is making more critical the necessity of lossy compression solutions able to achieve higher compression ratios (CRs). In this paper, a transform-based lossy compressor, namely lossy compression algorithm for hyperspectral image systems (HW-HyperLCA), is proposed as a modification of the original HyperLCA, specifically designed for being executed into space-grade field programmable gate array (FPGA) devices. The HW-HyperLCA compressor can be totally executed using integer arithmetic, using different levels of precision that can be adapted for achieving the best relation between compression performance and computational burden. The experiments carried out in this paper demonstrate that the HW-HyperLCA compressor is able to compress the hyperspectral images to less than 1 bit per pixel per band with low distortions (more than 40 dB of signal-to-noise ratio) while the performed complexity analysis demonstrates that it can be efficiently executed into space-grade FPGAs using very few computational resources and good results in terms of maximum clock frequency can be obtained.

This paper investigates surface temperature variables as they relate to passive microwave-derived surface freeze/thaw (FT) state and assesses the accuracy of such FT products relative to surface temperature. Utilizing retrievals from the soil moisture active/passive (SMAP), advanced microwave scanning radiometer, and special sensor microwave imager instruments, surface FT records have previously been derived globally. Moderate Resolution Imaging Spectroradiometer skin temperature, North American Land Data Assimilation System (NLDAS) skin, 0–10-cm soil layer, and 2-m air temperatures are compared to the various FT state products (FTSPs) by defining the threshold for FT state transitions at 0 °C. This paper utilizes the 2015–2016 overlap period in FT records within the NLDAS domain. Spatial variability of classification accuracy (CA) is then investigated over the study area. A proportional differencing method also enables the identification of biases between FTSPs and surface temperature variables. Additionally, by analyzing probability distribution functions of FTSPs associated with temperature values, we assess the distribution of temperature variables as they relate to FT classifications. Classification agreement is shown to vary with sensor configuration, seasonality, and retrieval time. Air temperature is found to have the highest CAs across FTSPs (81%–91%), while NLDAS soil exhibits a close relationship to FTSPs, especially in regard to SMAP products. Finally, ascending ( p.m. ) retrievals are shown to be increasingly linked to the selected temperature parameters as compared to descending ( a.m. ) observations. This paper contributes to an improved understanding of current FTSPs and will benefit efforts to enhance future FT products.

The Red River of the North basin (RRB) is vulnerable to spring snowmelt flooding because of its flat terrain, low permeability soils, and the presence of river ice jams resulting from the river's northward flow direction. The onset and magnitude of major flood events in the RRB have been very difficult to forecast, in part due to limited field observations of snow water equivalent (SWE). Coarse-resolution (25-km) passive microwave observations from satellite instruments are well suited for the monitoring of SWE. Despite routine use in the Earth sciences community to document the impacts of climate change, the use of passive microwave observations in operational flood forecasting is rare. This paper compares daily satellite passive microwave SWE observations from special sensor microwave/imager (SSM/I) and special sensor microwave imager/sounder (SSMIS), advanced microwave scanning radiometer for earth observing system (AMSR-E), and advanced microwave scanning radiometer 2 (AMSR2) from 2003 to 2016 to modeled output from the SNOw Data Assimilation System (SNODAS) and Global Snow Monitoring for Climate Research -2 (GlobSnow-2) in the RRB to determine the differences between the remotely sensed SWE estimates and the model products currently used in flood forecasting. Results show statistically significant differences between the satellite observations and SNODAS in the northern region of the basin that were not evident in the southern region. Satellite estimates of peak SWE in the forecast subbasins in the northern region were 40–125% higher than the model results which points to the lack of ground observations used to constrain the model simulations. This paper recommends that satellite SWE observations should be considered for improving operational snowmelt forecasting in the RRB.

This paper presents a novel registration method for oblique synthetic aperture radar (SAR) images based on complementary integrated filtering (CIF) and multilevel matching. Our algorithm is divided into three steps. First, we considered different type of noises and employed the CIF to increase the signal-to-noise ratio of SAR images. Second, complementary affine invariant features, namely maximally stable extremal regions and Harris-affine features, were extracted simultaneously from image pairs, and then the initial matches were obtained based on the scale invariant feature transform (SIFT) descriptor and Euclidean distance. Therefore, the fundamental and homography matrixes could be calculated between image pairs, and then more matches were obtained under quasi-affine invariant SIFT (QAISIFT) and the hybrid geometric constraints. We further implemented the least square matching (LSM) based on the second-polynomial geometric model (SPGM), and thus the matching error of each corresponding point can be compensated according to the optimal SPGM. Third, the precise registration was achieved based on the matches of the second step. Experiments on four groups of oblique SAR images demonstrated the effectiveness of the proposed method. The contribution of this paper includes three aspects. One is that the proposed CIF can remove SAR image noise as much as possible; another is that the proposed QAISIFT can achieve near affine invariance across viewpoint change images; the third is that the advanced LSM can notably improve the accuracy of feature matches.

Real-time hyperspectral imaging on-board compression represents a critical processing step in many remote sensing applications where the acquired hyperspectral data need to be efficiently stored and/or transferred. However, the complexity of the compression algorithms as well as the volume of data to be compressed and the limited computational resources of the hardware devices available on-board turn the real-time compression into a very challenging task. This paper presents a low-power-consumption solution for real-time lossy compression of hyperspectral images. The lossy compression algorithm for hyperspectral image system (HyperLCA) compressor has been implemented onto two NVIDIA Jetson developer kits. These NVIDIA boards include low-power embedded graphic processing units, which allow parallel programing for speeding up the compression process at a reasonable low power consumption. The experiments carried out in this paper are oriented to the necessities imposed by a specific smart farming application although all drawn conclusions are extrapolable to other fields in which remotely sensed hyperspectral images are to be compressed in real time. The obtained results verify both the good performance of the HyperLCA compressor for the targeted application and the achievement of a real-time performance by using the developed implementations. Additionally, several comparisons and conclusions have been drawn from the experiments in relation to the different strategies employed for accelerating the compression process.

In this paper, the development and implementation of the clustering algorithm (CA) developed for an instrument to detect lightning phenomena (Lightning Imager) is presented. The aim of the Lightning Imager (LI) instrument is to provide information relevant to the localization and the radiance of lightning events with respect to terrestrial systems. A CMOS sensor with $1170 \times 1000$ resolution and 1000 fps frame-rate is used on the instrument, making the LI able to identify smaller lightning in terms of dimension (minimum diameter 10 km) and temporal pulse duration (minimum 0, 6ms). The amount of acquired data is very large, but only a small subset of pixels contains useful lightning information, thus information clustering is fundamental for downlink data-rate reduction. Raw sensor data are processed by application-specific integrated circuits (ASIC) in order to extract pixels coordinates belonging to lightning flashes (detected transients), which are then directly sent to the CA implemented in a field-programmable gate array (FPGA). Detected transients must be processed as they come to the CA due to high throughput requirement. The CA objective is to define rectangular windows that enclose more detected transients (DTs) in order to send more compact information to Earth. The design presented in this paper was carried out through the development of a high-level model in order to verify algorithm functionality and compliance with performance requirements. A lightning event generator was developed to simulate the DTs coming from the ASIC part.The hardware architecture was conceived and a bit-true model was developed to evaluate the implementation loss.

Many applications of hyperspectral remote sensing involve automatic detection of solid targets. The design and applicability of such systems depend on their detection performance under various deployment scenarios; thus, it is useful to develop an accurate and realistic performance model. In this paper, we present a hyperspectral target detection performance prediction model for the common matched filter and normalized matched filter detection algorithms. A statistical model for hyperspectral data is discussed, which includes a replacement model for target pixels and a finite mixture of $t$ -elliptically contoured distributions for background materials. For this general input model, analytic forms for detector output distributions are not always available. The main contribution of this paper is to develop an efficient and robust performance simulation technique using Monte Carlo (MC) with importance sampling. Compared to standard MC, our simulation technique requires a substantially smaller sample size to obtain accurate performance estimates at low false alarm rates. The proposed technique is useful for predicting detection performance over a wide range of input models when analytic solutions are unavailable.

In this paper, a new framework for iterative speckle noise reduction in polarimetric synthetic aperture radar (SAR) data is introduced. Speckle is inherent to all coherent imaging systems and affects SAR imagery in the form of strong intensity variations in pixels with similar backscattering coefficients. This makes the interpretation of SAR data in several applications a difficult task. The proposed framework includes a preprocessing step capable of dealing with noise correlation usually found in single-look data. The general filtering approach is based on the Beltrami flow for denoising manifolds or images painted on manifolds. The principal contribution of this work is to adapt this approach to deal with covariance or coherency matrices instead of optical imagery. The evaluation presented suggests that this approach allows for good spatial and radiometric preservation compared to other state-of-the-art methods. Experiments are performed on the basis of synthetic and real-world experimental data. The validation of the proposed framework is accomplished using two refined error performance measures and the well-known effective number of looks measured. The source code of a parallel implementation of the proposed framework is released under the MPL 2.0 ( https://www.mozilla.org/en-US/MPL/2.0/ ) alongside this paper.

The Fluorescence Explorer (FLEX) is an Earth observation mission currently in development by the European Space Agency to perform quantitative measurements of the solar induce vegetation fluorescence. As a core of the mission concept, FLEX is planned to be launched by the end of 2023 and shall fly in tandem with one of the Copernicus Sentinel-3 satellites, which is already in orbit. This situation will allow FLEX to benefit from the optical and thermal sensors of Sentinel-3 and provide an integrated package of measurements. This paper presents the preliminary parametric analysis on the along-track dynamic of the tandem flight formation between Sentinel-3 and FLEX. It includes the study of the maneuvering strategy for a master–slave scenario concept, where Sentinel-3 acts as the master of the formation, whereas FLEX is the slave. This control strategy is assessed in terms of safety for both satellites and performance, where it is of interest to maintain the along-track distance between satellites as bounded as possible. Numerical and analytical results of this paper are provided showing that the control strategy proposed is able to fulfill the mission requirements under the expected conditions while providing enough time for the control center to coordinate the orbital maneuvers of Sentinel-3 and FLEX. Additionally, and from the results obtained in this paper, it is concluded that the safest configuration for the mission is locating FLEX ahead of Sentinel-3.

Opencast mining has huge effects on water pollution for several reasons. Fresh water is heavily used to process ore. Mine effluent and seepage from various mine related areas especially tailing reservoir, increase water pollution immensely. Monitoring and classification of mine water bodies, which have such environmental impacts, have several research challenges. In the past, land cover classification of a mining region detects mine and non mine water bodies simultaneously. Water bodies inside surface mines have different characteristics from other water bodies. In this paper, a novel method has been proposed to differentiate mine and non mine water bodies over the seasons, which does not require to set a threshold value manually. Here, water body regions are detected over the entire scene by any classical water body detection algorithm. Further, each water body is treated independently, and reflectance properties of a bounding box over each water body region are analyzed. In the past, there were efforts to use clay mineral ratio (CLM) to separate mine and non mine water bodies. In this paper, it has been observed that iron oxide ratio (IO) can also separate mine and non mine water bodies. The accuracy is observed to increase, if the difference of CLM and IO is used for segregation. The proposed algorithm separates these regions by taking into account seasonal variations. Means of differences of CLM and IO of each bounding box have been clustered using K -means clustering algorithm. The automation provides precision and recall for mine, and non mine water bodies as $[77.83\%,76.55\%]$ and $[75.18\%,75.84\%]$ , respectively, using ground truths from high-definition Google Earth images.

It is a challenge to achieve high accuracy altimetric results using space-based global navigation satellite system reflectometry (GNSS-R). This paper focuses on improving the accuracy of space-based GNSS-R altimetric measurements by considering receiver dynamics in the geometric model. The receiver dynamics has negligible impact on the geometric model of low-platform GNSS-R because the receiver movement is quite small during the time period between receiving direct and reflected signals. However, in the case of space-based GNSS-R, the receiver has much higher dynamics, thus the receiver dynamics cannot be ignored. This paper analyzes the altimetric error caused by the receiver dynamics in space-based GNSS-R. Theoretical studies and simulations were conducted to evaluate the magnitude of the altimetric error caused by receiver dynamics. The results show that the dynamics of a low earth orbit receiver can lead to errors of the order of several meters, depending on the trajectory angle of the receiver. The results using data from the TDS-1 mission conclude that neglecting receiver motion causes an average error of 3.4 m. This agrees with the result obtained from a theoretical analysis. Therefore the geometric model that takes into account receiver dynamics can improve the accuracy of space-based GNSS-R altimetric measurements of the order of a few meters compared to neglecting receiver dynamics.

When oceanic rainfall occurs, it creates a vertical salinity profile that is fresher at the surface. This freshwater lens is mixed downward by turbulent diffusion, dissipating over a few hours until the upper layer (1–5 m depth) becomes well mixed. Thus, there will be a transient bias between the in situ bulk salinity and the satellite-measured sea surface salinity (SSS) (representative of the first centimeter of the ocean depth). Based on measurements of Aquarius (AQ) SSS under rainy conditions, a model called rain impact model (RIM) was developed to assess the SSS variations due to the accumulation of rainfall prior to the time of the AQ observation. RIM uses ocean surface salinities from hybrid coordinate ocean model and the NOAA global precipitation product, climate prediction center morphing, to estimate changes in the near-surface salinity profile. Also, the RIM analysis has been applied to soil moisture and ocean salinity with similar results observed. The Soil Moisture Active Passive (SMAP) satellite carries an L -band radiometer, which measures SSS over a swath of 1000 km at 40-km resolution. SMAP can extend AQ salinity data record with improved temporal/spatial sampling. This paper describes RIM that simulates the effects of rain accumulation on SMAP SSS, showing good correlation between the model and the observed SSS values. Given the better resolution of SMAP, the goal of this paper is to continue the previous analysis of AQ to better understand the effects of the instantaneous and accumulated rain on the salinity measurements.

Authentic remote-sensing images suffer non-uniform complex distortions during acquisition, transmission, and storage. Clouds, light, and exposure also affect local quality. This paper constructs a usability-based subjective remote-sensing image dataset and gives a definition of usability for images with non-uniform distortion, where the image usability is determined by the weighted quality of image's blocks. It is difficult to extract the handcraft features from remote-sensing images with complex mixture distortion. Recently, convolutional neural network (CNN) has been introduced into blind quality assessment for images with uniform distortion, which includes feature learning and regression in one processing. In this paper, we first describe and systematically analyze the usability of remote-sensing images in detail. Then, we propose a remote-sensing image usability assessment (RSIUA) method based on a residual network by combining edge and texture maps. The score of remote-sensing image usability was obtained with the weighted averaging of the quality scores of all image blocks, and the weight of each image block was determined by its quality score. We compared the proposed method with three traditional image quality assessment methods, one CNN-based method for images with simulated distortion, and one scale-invariant feature transform-based RSIUA method. The linear correlation coefficient, Spearman's rank ordered correlation coefficient, and root-mean-squared error of experiments demonstrate that our method outperforms all five competitors. The experiments also reveal that the edge and texture maps can improve the performance.

Hyperspectral (HS) image sharpening (namely hyper sharpening) with an auxiliary sensor, such as multispectral (MS) or panchromatic sensor, has attracted a great deal of attention for the past decade. A number of hyper-sharpening techniques, aiming at enhancing the spatial resolution of HS images, have been developed and demonstrated their effectiveness especially on synthetic or simulated data. Nevertheless, since the differences between different remote sensing systems or imaging conditions are complicate, it results in a serious spectral distortion when applying on real remote sensing data acquired by different sensors under different acquisition times or conditions. Unfortunately, very few works have considered this issue. In this paper, a new hyper-sharpening framework based on spectral modulation is proposed to better preserve spectral information when fusing with MS data acquired by a different sensor. The goal of this paper is to generate an adjusted MS image that would have been observed under the same imaging conditions with the corresponding HS sensor. Two approaches, originating in MS pan-sharpening field, are introduced as examples under this framework, namely high-pass details injection model and band-dependent spatial-detail model. Experiments on three HS and MS datasets acquired by different platforms demonstrate that the proposed framework is beneficial to the spectral fidelity of the fused image compared with some state-of-the-art hyper-sharpening techniques. Meanwhile, it is also easy to implement and has a certain advantage in enhancing the spatial details.

This paper is devoted to detect and classify intra-urban changes by jointly exploiting Sentinel-1 (S-1) SAR data and nighttime light data. By extracting urban extents and urban density maps from SAR data, changes in nighttime lights can be used to detect changes related to the level of activity in a specific portion of each urban areas. At the same time, changes in radar backscattering are prone to reveal changes in the two- and three-dimensional structures of the built-up. The combination of these multimodal datasets has already proved to be useful to discriminate urban change patterns at the city level. In this paper, instead, SAR datasets from S-1 are exploited, allowing the recognition of different intra-urban changes. Experimental results focus on fast growing (mega) cities in East Asia, allowing us to understand in a more detailed way how they are changing and evolving in all three dimensions. Examples for Nanjing, Shanghai, and Guangzhou (China), Saigon (Vietnam), and Vientiane (Laos) are discussed to prove this statement.

Detection, classification, and monitoring of surface mining region from satellite images have vast challenges. Surface mining industry has huge effect on social, economical, ecological, and environmental welfare of a country. Machine learning techniques are used to capture the unique characteristics of surface mining region from satellite images. In opencast coal mines, raw coals stay exposed to the environment. Reflectance measures in certain wavelengths of such areas are distinguishable from other non-coal areas. These surface mining areas can be detected through satellite images using this cue. It has been found in this paper, that the spectral index derived from SWIR-I (1.566–1.651 $\mu \text{m}$ ) and SWIR-II (2.107–2.294 $\mu \text{m}$ ) bands of Landsat 8 exhibit distinctive features to detect high mineral areas of coal seams. Spectral plots of reflectance values of different land cover classes with respect to reflectance values of coal are found to be different. This paper presents a novel index to detect surface coal mines. Coal quarry and coal dump regions are considered as the true positive regions. The accuracy of the index has been validated with high-resolution Google Earth ground truth images and statistical measures. The robustness of the index is analyzed through seasonal variations. The average accuracy of the proposed method over the season is found to be ${86.24\%}$ . The proposed index can be further used for classification and area monitoring of opencast coal mine regions.

The data available in the world come in various modalities, such as audio, text, image, and video. Each data modality has different statistical properties. Understanding each modality, individually, and the relationship between the modalities is vital for a better understanding of the environment surrounding us. Multimodal learning models allow us to process and extract useful information from multimodal sources. For instance, image captioning and text-to-image synthesis are examples of multimodal learning, which require mapping between texts and images. In this paper, we introduce a research area that has never been explored by the remote sensing community, namely the synthesis of remote sensing images from text descriptions. More specifically, in this paper, we focus on exploiting ancient text descriptions of geographical areas, inherited from previous civilizations, to generate equivalent remote sensing images. From a methodological perspective, we propose to rely on generative adversarial networks (GANs) to convert the text descriptions into equivalent pixel values. GANs are a recently proposed class of generative models that formulate learning the distribution of a given dataset as an adversarial competition between two networks. The learned distribution is represented using the weights of a deep neural network and can be used to generate more samples. To fulfill the purpose of this paper, we collected satellite images and ancient texts to train the network. We present the interesting results obtained and propose various future research paths that we believe are important to further develop this new research area.

Leaf area index (LAI) is an important forest structural parameter in the process of photosynthesis. Most studies have investigated the estimation of forest LAI using the spectral information of optical remote sensing images or the height information from LiDAR data. This paper explored the estimation of forest LAI using canopy cover and forest height information extracted from stereo imagery acquired by cameras onboard an unmanned aerial vehicle (UAV). UAV remote sensing has gradually become practical in recent years. Two alternative methods were proposed to extract forest height information. The height indices method extracted forest height indices within each forest plot based on the vertical histogram of the canopy height model derived from stereo imagery, while the segmentation method characterized forest plots using tree numbers and average tree heights based on the individual tree segmentation. The results showed that canopy cover and forest height are complementary in the estimation of forest LAI no matter what method was used. The combined use of canopy cover and forest height information extracted by the segmentation method had a better estimation accuracy of forest LAI with R 2 = 0.833 and RMSE = 0.288. This paper demonstrated a new approach to predict forest LAI using UAV optical images.

This foreword starts with a brief historical overview of quantitative land remote sensing modeling and estimation from roughly 1970 to the present. The papers of this special issue are mainly derived from the Third National Forum on Quantitative Remote Sensing was held at Beijing Normal University held 14–15 July 2017. There were 296 meeting participants from 65 research institutes and universities in China, and almost all aspects of quantitative land remote sensing were discussed. There were 36 submissions, but only 13 papers were selected covering the following topics: modeling the effects of surface topography on directional shortwave reflection and longwave emission, radiometric calibration, and atmospheric correction, inversion algorithm for estimating downward longwave radiation, surface net radiation, forest leaf area index (LAI), winter wheat LAI and CWC, the Global LAnd Surface Satellite (GLASS) land surface temperature (LST) and fraction of vegetation cover (FVC) products, and evaluation and validation of satellite products of incident shortwave radiation, fractional snow cover (FSC), and cloud cover and cloud optical depth. Brief summaries of these papers are provided.

This paper presents an advanced urban traffic density estimation solution using the latest deep learning techniques to intelligently process ultrahigh-resolution traffic videos taken from an unmanned aerial vehicle (UAV). We first capture nearly an hour-long ultrahigh-resolution traffic video at five busy road intersections of a modern megacity by flying a UAV during the rush hours. We then randomly sampled over 17 K 512×512 pixel image patches from the video frames and manually annotated over 64 K vehicles to form a dataset for this paper, which will also be made available to the research community for research purposes. Our innovative urban traffics analysis solution consists of an advanced deep neural network (DNN) based vehicle detection and localization, type (car, bus, and truck) recognition, tracking, and vehicle counting over time. We will present extensive experimental results to demonstrate the effectiveness of our solution. We will show that our enhanced single shot multibox detector (Enhanced-SSD) outperforms other DNN-based techniques and that deep learning techniques are more effective than traditional computer vision techniques in traffic video analysis. We will also show that ultrahigh-resolution video provides more information that enables more accurate vehicle detection and recognition than lower resolution contents. This paper not only demonstrates the advantages of using the latest technological advancements (ultrahigh-resolution video and UAV), but also provides an advanced DNN-based solution for exploiting these technological advancements for urban traffic density estimation.

A study investigating the feasibility of retrieving wind direction using low-altitude backward geometry of Global Navigation Satellite System-Reflectometry (GNSS-R) is implemented in this paper. This paper first focuses on analyzing the influence of wind direction on the roughness of sea surface and scattering strength. The results show that backward scattered GNSS signals obviously change over wind direction despite of the weaker power level than the forward signals. Then, the power link of backward scattered GNSS signal and the visible number of the satellites composing the bistatic backward observation geometry with the receiver are analyzed. Backward scattered GNSS signals are lower about $10\text{--}20$ dB than forward ones; hence, a higher gain left-hand circularly polarization antenna should be used to receive them. In the same antenna view, theoretically the backward scattered signals from about three satellites could be received. Finally, a retrieving algorithm of wind direction based on matching theory is proposed. The retrieving results using simulated data show that, first, multibeam observation (at least three-beam observation) should be utilized to remove the uncertainty of retrieving wind direction, second, the accuracy of wind speed and the signal-to-noise ratio (SNR) of the delay waveform importantly impact on the retrieving performance of wind direction, therefore, it is needed to improve the measuring accuracy of wind speed and the SNR. In one word, all results mentioned above give the guide information to observe wind direction using low-altitude GNSS-R.

Anomaly detection is one of the practical applications in hyperspectral imagery (HSI) over the last two decades. In this paper, we propose a combined similarity criterion anomaly detector (CSCAD) method for HSI anomaly detection. The proposed method approximates the background using the surrounding neighbor pixels. Then, the residual error is obtained by subtracting the approximated background from the original data. Similarity criterion between the central and the adjacent pixels is applied simultaneously to detect anomalies more precisely. Finally, the criterion number is calculated for each pair of the central-adjacent pixels in the dual window, and then anomalies are detected. The main contribution of the paper is removing the noise and outlier, employing an efficient method that conducts us to more accurate detection performance. Contamination of the noise and outlier is reduced by the residual error and similarity criterion values. So, applying the criterion numbers is an accurate method for removing the outliers and noise without any need to use an additive stage that was introduced in our previous work. We implement the proposed method on three real hyperspectral images. CSCAD outputs are depicted by receiver operating characteristic curves, area under curve values, and intuitive images. Comparing the results of the proposed method with eight popular and the state-of-the-art methods proves that the CSCAD is an accurate and effective method to detect anomalies.

The papers in this special section were presented at the Pattern Recognition in Remote Sensing (PRRS) Workshop. The technical committee for pattern recognition in remote sensing (PRRS) and mapping (TC7) of the international association for pattern recognition (IAPR) organizes a biennial workshop on PRRS. This workshop series has been a popular forum for experts of both communities, pattern recognition and remote sensing, and accordingly there is co-sponsoring from IAPR, IEEE-GRSS, and ISPRS. In particular, the intercommission working group for pattern analysis in remote sensing (ICWG II/III) of the ISPRS is closely cooperating. The latest workshops have been very successfully held in December 2016 in Cancun, Mexico, and in August 2018 in Beijing, China. Additionally, TC7 compiles special issues on PRRS in journals such as IEEE-JSTARS or Pattern Recognition Letters [items 1)-6) in Related Works]. The issue at hand fits in this series. Contained are 29 papers, including some work of authors that have participated and published preceding work in the 2016 PRRS in Cancun.

Conventional scatterometer wind retrieval has been proved effective for low and moderate wind conditions. Under high wind conditions, especially for typhoons, the air–sea interactions are fierce, and usually the sea surface backscattering is complicated by rain, leading to degraded scatterometer wind quality. Under such circumstances, radiometers in different frequencies can provide information of rains and the complex air–sea interface. In this paper, an artificial neural network (ANN) was employed to combine the observations of the two sensors onboard HY-2A satellite: radiometer and scatterometer (HSCAT) to achieve wind field retrieving under typhoon conditions. The ANN was trained by using global/regional assimilation and prediction system (GRAPES) wind field data as true values. The established network is then verified by an independent dataset excluded from the training data. It is shown that the wind speeds and directions retrieved from the ANN are better than those of the operational HSCAT products as compared to GRAPES winds. Further comparison with the H*Wind product proves that the proposed ANN is effective in terms of high wind field retrieval. The research of this paper provides a nice reference for the typhoon wind retrieval from the HY-2A satellite and for the data processing of the coming HY-2 satellite series.

The safety of the electricity infrastructure significantly affects both our daily life and industrial activities. Timely and accurate monitoring of the safety of electricity network can prevent dangerous situations effectively. Thus, we, in this paper, develop a voxel-based method for automatically extracting the transmission lines from airborne LiDAR point cloud data. The method proposed in this paper uses three-dimensional (3-D) voxels as primitives and consist of the following steps: First, skeleton structure extraction using Laplacian smoothing; second, feature construction of a 3-D voxel using Latent Dirichlet allocation topic model; and third Markov random field model-based extraction for generating locally continuous and globally optimal results. To evaluate the effectiveness and robustness of the proposed method, experiments were conducted on four different types of power line scenes with flat and complex terrains from helicopter-borne LiDAR point cloud data. Experimental results demonstrate that our proposed method is efficient and robust for automatically detecting both the single conductor and the bundled conductors, with precision, recall, and quality of over 96.78%, 98.67%, and 96.66%, respectively. Moreover, compared with other existing methods, our proposed method provides higher detection correctness rate.

To perform better monitoring of infrastructure stability in urban and suburban regions, we propose an improved method based on the combined analysis of persistent scatterer (PS) and distributed scatterer (DS). Previous work [13] is extended to explore the DS measurements by exploiting an adaptive homogeneous filter and the capon-based estimator. In this paper, PSs are detected as reference points in the first-tier network. Parameters along network arcs are estimated through the integrated use of beam-forming and an M-estimator. In the second-tier network, we design an adaptive homogeneous filter to cluster statistically homogeneous pixels. DS candidates are then connected to their nearest PS references, establishing the DS network. We estimate DS parameters by using capon beamforming. As the proposed method can provide more complete deformation details of low-coherence targets, it is more effective in stability monitoring. Results from 28 C-band ENVISAT-ASAR scenes of the Hong Kong International Airport are presented in this paper.

In this paper, we describe our advances in manufacturing a 256-layer 7- μ m thick harmonic lens with 150 and 300 mm focal distances combined with color correction, deconvolution, and a feedforwarding deep learning neural network capable of producing images approaching photographic visual quality. While reconstruction of images taken with diffractive optics was presented in previous works, this paper is the first to use deep neural networks during the restoration step. The level of imaging quality we achieved with our imaging system can facilitate the emergence of ultralightweight remote sensing cameras for nano- and pico-satellites, and for aerial remote sensing systems onboard small UAVs and solar-powered airplanes.

Recent years have seen a surging interest in several novel tools pertaining to SAR, such as multichannel high-resolution and wide-swath (HRWS) SAR, multibaseline interferometric SAR (InSAR), multisubband SAR, polarimetric SAR (PolSAR), and polarimetric SAR interferometry (PolInSAR). We believe that these new approaches to SAR have valuable scientific applications. Here, we present a new experimental airborne SAR system named “N-SAR” (SAR of the Nanjing Research Institute of Electronic Technology) that can fulfill new requirements and is scalable to allow rapid development of modern SARs. As a dual-antenna airborne SAR system, the N-SAR system will be used to test new technologies and signal processing algorithms such as PolSAR and PolInSAR, multibaseline SAR interferometry, multichannel multichannel HRWS SAR/InSAR, and SAR ground moving target indicator. It will play a key role in evaluating the performance of current engineering-oriented SAR systems by using several new operational modes for scientific purposes. In this paper, we provide a conceptual description of the general system design features, instrument design, and capabilities of the N-SAR system. To meet the requirements of different experiments, a novel operational mode, named the alternating bistatic multipolarized mode based on the N-SAR system, is presented. A series of flight tests that started in April 2017 and will be carried out over the next few years are shown. Several preliminary experimental results pertaining to the calibration of PolSAR, multichannel SAR imaging, and interferometry are presented in this paper as an early validation of the capabilities of the N-SAR system.

This paper presents a new method for quantifying the carbon storage in urban trees using multispectral airborne laser scanning (ALS) data. This method takes the full advantage of multispectral ALS range and intensity data and shows the feasibility of quantifies the carbon storage in urban trees. Our method consists of four steps: multispectral ALS data processing, vegetation isolation, dendrometric parameters estimation, and carbon storage modeling. Our results suggest that ALS-based dendrometric parameter estimation and allometric models can yield consistent performance and accurate estimation. Citywide carbon storage estimation is derived in this paper for the Town of Whitchurch–Stouffville, Ontario, Canada, by extrapolating the values within the study area to the entire city based on the specific proportion of each land cover type. The proposed method reveals the potential of multispectral ALS data in land cover mapping and carbon storage estimation at individual-tree level.

This paper describes spectral stray light measurements of diode array spectrometers and, in particular, the dependence of stray light and bandpass properties on the illumination beam geometry. The effects of underfilling and overfilling the nominal field of view of the instrument are presented. Our measurements showed that the line spread function of a commercially available miniature spectrometer module having a fiber optic entrance with 0.22 nominal numerical aperture can depend on fiber illumination cone apex angle up to $40^\circ$ , which is well above the $25^\circ$ nominal acceptance cone apex angle of the fiber. The bandpass width and amount of stray light inside the spectrometer increased proportionally with the illumination cone angle. Spectral stray light measurements with different setups gave consistent and repeatable results for similar illumination geometries. Care should be taken to ensure that illumination geometry during spectral stray light characterization of a spectrometer matches the conditions during target measurements for which the spectral stray light corrections are applied. The impact of illumination geometry during stray light characterization depends on the target signal and particular application. Photochemical reflectance index (PRI) and normalized difference vegetation index (NDVI) of a hemiboreal birch forest are used as contrasting examples in this paper. Compared to natural variability, the relative impact on PRI exceeds that of NDVI by 10 times.

Standard deviation (SD) is one of the most famous features for image processing. It is widely used for ship detection in synthetic aperture radar imagery. However, it is very sensitive to the area ratio of target. It cannot reflect intensity fluctuation appropriately in some cases. Traditional constant false alarm rate (CFAR) detectors, e.g., cell-averaging CFAR, fail in dense target situations. This paper shows that this classical problem can be understood as a same problem in terms of area ratio variant and area ratio invariant (ARI). It reinterprets the reason for CFAR performance degradation and provides new insight into this problem. This paper begins with the proposal of an ARI-SD feature. Then, ARI-feature group (ARI-FG) is defined. Finally, traditional CFAR detectors are modified by ARI-FG. Compared to traditional features, the proposed ARI features are independent of the area ratio, by means of an elimination of the influence from area variation. Experimental results based on real data show that the modified CFAR method does not cause detection loss and additional false alarms compared to traditional CFAR. It is more effective and robust in dense target situations. In addition, the potential of ARI-FG for ship discrimination is also explored in this paper.

This paper analyzes and compares multiple techniques for estimating the mean square slope (MSS) of surface waves during Hurricane Ike in the Gulf of Mexico and studies the correlation of the MSS estimates with wind speed measurements along the same tracks. Three separate instruments collected measurements in parallel, including a GPS reflectometry (GPS-R) receiver, a stepped frequency microwave radiometer (SFMR), and a wide swath radar altimeter (WSRA). These datasets were used to study the correlation between the flight level and near-surface wind and MSS during Hurricane Ike in 2008. The GPS-R, SFMR, and WSRA instruments recorded temporally and spatially coincident data during two passes over the hurricane eye. This paper estimates the ocean surface MSS using GPS-R for two eye transects using: a least squares model fitting technique, the reflected signal waveform width, and an integration of the reflected signals in an area around the peak. Subsequently, the correlations between the GPS-R and WSRA MSS estimates and the SFMR wind speed estimates are compared to reveal regions of high and low surface MSS to wind speed correlation. Finally, a relationship between wind and MSS was derived from the GPS-R and SFMR data and is compared to existing MSS/wind models, including the results obtained by Katzberg et al. for hurricane conditions.

Mathematically, once the measured radiance has been corrected for atmospheric effects, the only issue when determining the land surface temperature and emissivity (LST and LSE) is solving the ill-posed problem in the radiative transfer equation (RTE). Recently, based on the wavelet transform theory, a so-called wavelet-transformed temperature and emissivity separation (WTTES) method has been proposed for retrieving LST and LSE from hyperspectral data. Although, in a previous article, an initial suggestion was provided after analyzing the uncertainties under the conditions of several typical errors, it was also noted that considerable work was still necessary for achieving a reliable method for driving the WTTES algorithm, particularly under different situations. To complement the previous analysis of the WTTES algorithm, this paper presents a more detailed and comprehensive evaluation in which we changed the wavelet functions, varied the wavelet levels, and biased the atmospheric profiles. The results in this paper showed that the WTTES algorithm was insensitive to the choice of wavelet function. In addition, the WTTES algorithm could stay stable in most circumstances. A wavelet level of n = 3 was more recommend when the NEΔT was approximately 0.2 K. When a higher level of noise was found, a level of n = 4 could be then used to better overcome the noise. When a lower level of noise was found, a level of n = 2 could be used to further refine the spectral features. Additionally, we also found that the WTTES algorithm could have problems when atmospheric effects were inaccurately compensated for, especially for wet-warm profiles.

This paper examines the radio frequency interference (RFI) environment within microwave imager bands for sensors that have observed the earth over the last two decades. Since the microwave imagers have used various bands both within and outside of International Telecommunications Union (ITU) allocated frequencies for passive satellite earth exploration, this survey provides valuable insight into band selection and mitigation strategies for future missions. Several conclusions are drawn from this paper. First, significant land-based RFI exists at the C -Band. The two-band mitigation solution for the C -Band utilized by Advanced Microwave Scanning Radiometer 2 provides some RFI reduction of high-level RFI (>10 K), and therefore, may have merit for future missions in cases where only two subbands can be feasibly implemented. At the X -band, ocean reflections from direct broadcast and communication satellites (especially around Europe) provide considerable interference above the ITU-allocated 10.6–10.7 GHz band, whereas very little reflected interference is observed within the allocated band. Good out-of-band rejection is required to avoid the reflected RFI above 10.7 GHz. Both inside and outside of the X -band allocation, significant terrestrial RFI exists. For the K -band, the 19.35-GHz band utilized by Special Sensor Microwave Imager Sounder and Tropical Rainfall Measuring Mission Microwave Imager avoids reflected RFI from satellites that is present around the continental United States for satellites observing within the 18.6 to 18.8 GHz allocated band. The analysis suggests that the 19.35-GHz band may be preferable to the 18.7 GHz allocated band for ocean applications if avoiding RFI is desired.

This paper describes the technical approach to generate homogeneous warm brightness temperature scene binary mask over the earth for the intercalibration of microwave imagers. The objective is to identify homogeneous desert scenes as targets in radiometric intercalibration of multisource multisensor systems such as Global Precipitation Measurement (GPM) constellation. In this paper, a method is developed to generate mask over continental Australia using 10, 18, and 37 GHz vertical and horizontal polarized radiance observations from GPM Microwave Imager data. Major factors considered are diurnal variation, second Stokes parameter, spatial homogeneity, and surface emissivity. Major advantage of this tool is its minimal, almost negligible, dependence on any external ancillary modeled or observed data.

Low-cost unmanned aerial vehicles (UAVs) utilizing push-broom hyperspectral scanners are poised to become a popular alternative to conventional remote sensing platforms such as manned aircraft and satellites. In order to employ this emerging technology in fields such as high-throughput phenotyping and precision agriculture, direct georeferencing of hyperspectral data using onboard integrated global navigation satellite systems (GNSSs) and inertial navigation systems (INSs) is required. Directly deriving the scanner position and orientation requires the spatial and rotational relationship between the coordinate systems of the GNSS/INS and hyperspectral scanner to be measured. The spatial offset (lever arm) between the scanner and GNSS/INS unit can be measured manually. However, the angular relationship (boresight angles) between the scanner and GNSS/INS coordinate systems, which is more critical for accurate generation of georeferenced products, is difficult to establish. This paper presents three calibration approaches to estimate the boresight angles relating hyperspectral push-broom scanner and GNSS/INS coordinate systems. For reliable/practical estimation of the boresight angles, this paper starts with establishing the optimal/minimal flight and control/tie point configuration through a bias impact analysis starting from the point positioning equation. Then, an approximate calibration procedure utilizing tie points in overlapping scenes is presented after making some assumptions about the flight trajectory and topography of covered terrain. Next, two rigorous approaches are introduced – one using ground control points and other using tie features. The approximate/rigorous approaches are based on enforcing the collinearity and coplanarity of the light rays connecting the perspective centers of the imaging scanner, object point, and the respective image points. To evaluate the accuracy of the proposed approaches, estimated boresight angles are used for orthorectification of six hyperspectral UAV dataset acquired over an agricultural field. Qualitative and quantitative evaluations of the results have shown significant improvement in the derived orthophotos to a level equivalent to the ground sampling distance of the used scanner (namely, 3–5 cm when flying at 60 m).

With the successful launch of GaoFen4 (GF4), on-orbit high accuracy geometric calibration for the high-resolution geostationary optical satellite will be a new research topic. With the improvement in the geometric resolution from geostationary orbit, it will become more and more difficult to meet the requirements of both high geometric resolution and large coverage for the available reference data. The purpose of this paper is to explore a new self-calibration mode for GF4 and future high-resolution geostationary optical area array cameras based on the fewest ground control points (GCPs). To overcome the problems of overparameterization, strong correlation and lower significance of the traditional rigorous imaging model, the simplified physical internal model is proposed, and its effectiveness in describing and compensating for the camera internal distortion is verified. Based on the simplified physical internal model, the self-calibration method based on two GCPs and evenly distributed tie points of two images is proposed for the high accuracy estimation of the calibration parameters. The GCPs can be used to provide the absolute geographical constraints for scale information, and the tie points can be used to provide the global constraints for optimum estimation. After calibration, the internal distortion is well compensated, and the positioning accuracy with relatively few GCPs is shown to be better than 1.0 pixels for both the panchromatic and near-infrared sensor and the intermediate infrared sensor. This paper will provide a new usable concept and approach for the future higher resolution geostationary area array optical camera to overcome the stringent requirements of both high resolution and a large area of reference data for the traditional calibration method.

This paper presents the first results on comparisons of Scatterometer Satellite-1 (SCATSat-1) derived wind datasets with the in situ , reanalysis as well as another operational scatterometer derived winds in the Bay of Bengal during the period November 2016–March 2017. The comparisons of daily gridded wind products of SCATSat-1 with buoys show good correlations (>0.83), higher skill scores (>0.92), and lower root mean square errors (RMSEs) of 0–2 m/s for wind speeds (WS) at the buoy locations. Similarly, the results corresponding to wind directions (WD) show higher correlations (>0.95), higher skill scores (>0.96), and relatively lower RMSEs (15–30°). Further, the intercomparisons of SCATSat-1 with Advanced Scatterometer and European Centre for Medium Range Weather Forecasts reanalysis winds show strong correlations for both WS (>0.85) and WD (>0.94). This paper also reports the capability of SCATSat-1 to capture three tropical cyclones Kyant, Vardah, and Mora during the period of study with the highest WS of 23.5 m/s.

The constant false alarm rate with convolution and pooling (CP-CFAR) method, which can improve the detection efficiency via GPU parallel acceleration in the airborne synthetic aperture radar (SAR) images, is proposed in this paper. The constant false alarm rate (CFAR) method is one of the most widely used methods for target detection in airborne SAR images. However, since the CFAR is performed on a pixel-by-pixel basis, the time consumption will increase rapidly with the expansion of image scene. Even if the GPU is used for acceleration, the efficiency improvement is still limited, which cannot meet the real-time processing requirements. Therefore, the CP-CFAR method for the target detection of SAR images is proposed in this paper. The convolution layer uses the horizontal and vertical Sobel operators to improve the contrast between targets and background, and the pooling layer can reduce the processing dimension of the images. The convolution and pooling layers are added before the two-parameter CFAR, which can reduce the computational elements but without losing the main feature of the original image. More importantly, compared to the traditional CFAR, the proposed CP-CFAR is more suitable for GPU acceleration, which can improve the detection efficiency significantly. Experiments on the moving and stationary target acquisition and recognition SAR images with a size of 1478 × 1784 show that, compared with the traditional cell-averaging CFAR, two-parameter CFAR and their CPU, multithread CPU, and GPU acceleration modes, the proposed CP-CFAR with GPU acceleration can obtain the best detection performance with the highest acceleration ratio, and the operation time is less than 192 ms.

Since the all-day and all-weather working characteristics, polarimetric synthetic aperture radar (PolSAR) has been widely used in many application for earth observation. PolSAR system can acquire targets scattering information under different polarization patterns and obtain detailed information, such as physical characteristics, spatial distribution and direction of the targets, etc. However, in high-resolution PolSAR images, the scattering characteristics of targets are very sophisticated, and different parts of the same target may present different scattering mechanisms. As a result, traditional interpretation technologies are sometimes not applicable in a high-resolution PolSAR scenario. Recently, many artificial intelligence algorithms that simulate the scheme of human cognition have been developed and they have great significance for efficient, intelligent, and accurate images interpretation. In this paper, based on human cognition theory and taking into account the characteristics of targets in PolSAR images, an algorithm for the interpretation of various targets in high-resolution PolSAR images is presented. In this paper, a “perception–reasoning–decision” cognition model is built following the cognition process of professional PolSAR image interpreters. In addition, with the help of theories such as multilevel image segmentation, fuzzy logic, neural network, and the context semantic characteristics, the mathematical model is established and implemented. The proposed method is performed on several sets of PolSAR images obtained under different imaging conditions and is validated by comparing with existing classification methods. The results show that the proposed method can effectively improve the identification of targets from PolSAR images.

This paper deals with the problem of fast focusing the spotlight synthetic aperture radar (SAR) data in high-resolution and high-squint mode, which has shown its potential capacity in various military SAR applications. Fast factorized backprojection (FFBP) is considered as an ideal methodology for high-resolution and high-squint SAR imagery. However, FFBP performs pixel-by-pixel interpolation, yielding heavy computational burden and limiting its real-time processing. In this paper, target detection technique, which is used as a discrimination tool to retain the target pixels and reject the clutter and noise pixels, is integrated into the subaperture imaging chain during FFBP implementation, aiming to speed up the overall image formation time. Due to the characteristic of spotlight acquisition, an adaptive processing scheme is developed for multiple subaperture images detection. With the integrated target detection technique, only the target pixels need to be interpolated and accumulated coherently. Without loss of focusing performance, the number of interpolations is dramatically reduced. The proposed fast imaging algorithm is named as adaptive FFBP (AFFBP), and it has distinctive superiority for sparse scene reconstruction, such as the maritime target imaging. AFFBP has been successfully applied to two simulated scenes with high and low signal-to-noise ratios (SNRs), and two real measured datasets with squint angle high up to $50^{\circ }$ and $70^{\circ }$ , respectively. All these experiments demonstrate the effectiveness and efficiency improvement of the proposed AFFBP algorithm.

This paper proposes a new framework of dictionary learning for a recently developed study, sparse representation of monogenic signal. The proposed framework is applied to target recognition in SAR image. Unlike the preceding works, where the sparse model is formed via an overcomplete dictionary whose atoms are the training samples themselves, a new approach to learn a more discriminative dictionary is proposed. To achieve target classification, two specific implementation schemes, global learning, and local learning, are developed. The global learning generates a single, universal dictionary for all target class. Since the class membership committed to each atom (element) is lost, the reconstruction error committed to each class is incapable to be computed. The conventional decision rule by the minimal residual could not be applied any more. Hence, this paper develops two decision rules for global learning. The first resorts to a third-party classifier, while the other recommends a nonparametric criterion. Different from the global learning, the local learning learns a subdictionary for each target class. These subdictionaries are concatenated to form a global one. Each subdictionary has been learned within certain target class, the generated atoms can be therefore labeled. The identification is predicted by evaluating which class of atom could produce the minimal reconstruction error. The effectiveness of proposed scheme is verified with multiple comparative studies.

In this paper, the authors propose a novel radio frequency interference (RFI) detection and removal method for a synthetic aperture radar (SAR). The proposed method is exclusively sensitive to a pulse-like signal that is temporally transmitted in wide bandwidth. The bandwidth of the RFI is supposed to be similar or wider than the bandwidth of the SAR, while traditional RFI detectors assume that it is narrower and thus it can be removed in the range-frequency and azimuth-time domain using a bandpass filter. On the other hand, the proposed method employs a local autocorrelation-based detection scheme in the range-time and azimuth-frequency domain. The proposed method assumes the existence of another type of RFI, namely, intermittently transmitted wideband (ITWB) RFI. Once ITWB RFI superimposes on the SAR image, its short-pulse duration/wide bandwidth hinders the detection in the range-frequency and azimuth-time domain. Conversely, in the range-time and azimuth-frequency domain, ITWB RFI can be detected easily because its Doppler frequency profile does not fit with backscattered signals from the ground. In this paper, a basic methodology and the experimental results are presented.

This paper proposes a diffusion filter based scheme to denoise seismic attributes and to improve the porosity volume, which is predicted from seismic attributes. We compare the performances of multiple diffusion [such as Perona–Malik diffusion filter, complex diffusion filter, improved complex adaptive diffusion filter (ICADF)] and nondiffusion (such as two-dimensional (2-D) median, 3-D median, smoothing, and bilateral filter) based filters in terms of four metrics such as root mean square error (RMSE), normalized RMSE, signal to noise ratio (SNR), and peak SNR (PSNR). In our earlier publication, we used an artificial neural network (ANN) to predict a lithological property (sand fraction) over a study area. We trained the ANN using an integrated dataset of low-resolution seismic attributes and a limited number of high-resolution well logs. In this paper, we generate the porosity volume from the seismic attributes using an ANN. The predicted porosity logs contain irregularities and artifacts due to the nonlinear mapping of the learning algorithm (e.g., ANN). We apply a set of filters to the output of the ANN to regularize the predicted porosity volume. The filtered porosity logs are compared with the generated log. The ICADF has been found to be most suitable for denoising the seismic data and the porosity volume. Generation of porosity maps from seismic inputs would be helpful to petroleum engineers for reservoir characterization.

The aim of this paper is to compare the potential of two popular flexible laws, the Fisher distribution and the Generalized Gamma distribution, for the statistical modeling of high-resolution SAR data through an original “mimicking-based” approach. The presented study allows to evaluate the ability of both laws to correctly imitate or “mimic” another reference law, frequently used for modeling the intensity of SAR images and chosen for instance as the $\mathcal {K}$ law or the Weibull, Beta or log-normal laws in this paper. This study uses log-cumulant statistics for parameter estimation of the imitating law and involves quantitative criteria of comparison based on the Kullback–Leibler divergences between the reference law and the Fisher law or the Generalized Gamma law. The mimicking capacities of both distributions are first analyzed for some sets of parameters describing different studied cases, covering a wide set of possible mimicked reference laws. The high modeling potential of both distributions is then illustrated on heterogeneous subscenes from real SAR intensity data. Pragmatical considerations are also taken into account to draw up recommendations about the preferential use of a distribution and to highlight complementarities of both Fisher and Generalized Gamma distributions, along with limitations of the approach.