Abstracts:With the dramatic growth of seismic data volume, efficient and accurate seismic source location has become a significant challenge to seismologists. Recently, time reversal imaging (TRI) has been widely applied in automatic seismic source location for its robustness and accuracy, but its wave-equation-based implementation is usually computationally expensive. To achieve an efficient <italic>in situ</italic> and real-time source location, the emerging sensor network is a good option. In this article, we propose a simplified Gaussian beam TRI (SGTRI) method to implement the seismic source location in a distributed sensor network. Gaussian beam (GB) is a high-frequency asymptotic solution of the wave equation, which can help reduce the computation costs of the wavefield extrapolation in conventional TRI. Traditionally, the GB construction for reflection seismic imaging covers the entire subsurface space. However, for certain source localization, only limited areas contribute. Thus, we propose a beamforming-technique-based simplified GB construction to further boost efficiency. Then, we propose an imaging condition for the SGTRI to construct the final source location map. Using synthetic experiments, we demonstrate the accuracy, robustness, and efficiency of the proposed method compared with conventional TRI. In the end, a field application also shows promising results.

Abstracts:To avoid decorrelation, conventional synthetic aperture radar interferometry (InSAR) requires that interferometric images should have a common spectral band and the same resolution after proper preprocessing. For a high-resolution (HR) image and a low-resolution (LR) one, the interferogram quality is limited by the LR one since the non-common band (NCB) between two images is usually discarded. In this article, we try to establish an InSAR method to improve interferogram quality by means of exploiting the NCB. To this end, we first define a new interferogram, which has the same resolution as the HR image. Then we formulate the interferometric relationship between the two images into a compressive sensing (CS) model, which contains the proposed HR interferogram. With the sparsity of interferogram in appropriate domains, we model the interferogram formation as a typical sparse recovery problem. Due to the speckle effect in coherent radar imaging, the sensing matrix of our CS model is inherently random. We theoretically prove that the sensing matrix satisfies restricted isometry property, and thus the interferogram recovery performance is guaranteed. Furthermore, we provide a fast interferogram formation algorithm by exploiting computationally efficient structures of the sensing matrix. Numerical experiments show that the proposed method provides better interferogram quality in the sense of reduced phase noise and obtain extrapolated interferogram spectra with respect to CB processing.

Abstracts:Remote sensing measurements of Earth materials are always made at scales much larger than individual grains and cavities, yielding only upscaled effective properties. An &#x201C;effective medium&#x201D; is an idealized uniform material that has the same measured properties as the real mixture. A uniform electric field applied to the ideal effective medium remains uniform within the sample; however, the same electric field applied to the composite results in fine-scale spatial variations of field strength within the sample, which depend on the properties of the constituents, their volume fractions, and their microgeometries. We derived analytic expressions for the electric field strength heterogeneity implicit in commonly used dielectric effective medium models. Only two-phase, statistically isotropic, low-loss materials, e.g., ice, snow, minerals, and freshwater in the microwave UHF band are considered. The method applies to singly or biconnected phases. The results confirm the uniform field in the isolated phase of material lying on the Hashin&#x2013;Shtrikman (HS) bounds; the continuous phase field variance increases with a decreasing volume fraction, approaching a well-defined limit as the fraction becomes vanishingly small. Expressions are found for field variance in higher-order composites of coated spheres, providing realizations of composites lying between the HS bounds, and illustrating field nonuniqueness when microstructure is unknown. The mean and variance of the field strength in popular effective medium models are also examined. Not only do the effective properties predicted by these models differ so do the electric field strength spatial variability, especially when the volume fraction of inclusions increases.

Abstracts:The new Ku-band scatterometer (HSCAT-B) onboard the HY-2B satellite was launched on October 25, 2018, and soon after the C-band scatterometer (Advanced Scatterometer (ASCAT)-C) onboard the MetOp-C satellite was launched on November 6, 2018. This article aims to validate the new sea surface wind products from them, and also to summarize the common issues in current scatterometer wind products. Thus, other scatterometer data are also used for comparisons, including the C-band MetOp-B/ASCAT, and Ku-band SCATSAT-1/OSCAT2 and HY-2A/SCAT winds. In this study, the C-band and Ku-band scatterometer wind products were each reproduced using the same procedures, in terms of backscatter calibration, wind retrieval, and quality control. The scatterometer winds are compared to the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5 winds or buoy winds, and the results show that the quality of ASCAT-C winds is almost the same as the well-known ASCAT-B; the HSCAT-B winds show quite good quality and similar validating statistics as ASCAT winds. Noticeable wind-speed-dependent biases are found in all Ku-band scatterometer winds, which suggests that refinements are needed for the NSCAT-4 geophysical model function, especially in terms of wind speed dependence for all incidence angles.

Abstracts:Shoreline location plays a key role in coastal research, management, and engineering. Remote sensing enables the quantification of shoreline information with the large spatial extent and high temporal frequency. Driven by river discharge and ocean dynamics, muddy coasts exhibit complicated spatiotemporal variations. It is essential, yet challenging to extract effective shoreline features from satellite images. Taking the Yellow River Delta coast in China as the study area, we present an indicator and min-cost approach to extract the shoreline in muddy coasts. The shoreline is represented as a set of linearly connected central points with high shoreline probabilities, and a set of image and spatial indicators are developed to assess these probabilities. The Salient Value indicator integrates the gradient magnitude and the edge intensity to detect the boundary strength; the Regional Difference indicator separates the water/land class from edge intensity to measure the possibility of being water or land; and the Seaward Distance indicator spatially distinguishes the true shoreline from other spectrally similar boundaries. A cost function combines these indicators to evaluate the local shoreline possibilities. A shoreline set is produced by an improved min-cost path method to evaluate the overall shoreline possibilities. The optimal shoreline paths are selected based on the parameter analyses of the shoreline set. The performance of the approach is confirmed by comparing with the ground truth and state-of-the-art methods. The effectiveness of the approach is tested for different spatial resolution data and coastal environments.

Abstracts:Object detection in remote sensing imagery is a critical yet challenging task in the field of computer vision due to the bird&#x2019;s-eye-view perspective. Although existing object detection approaches in remote sensing imagery have achieved great advances through the utilization of deep features or rotation proposals, but they give insufficient consideration to multilevel semantic information and its propagation for guiding the learning process. Accordingly, in this article, we propose a hierarchical semantic propagation (HSP) framework to boost object detection performance in remote sensing imagery, which is better able to propagate hierarchical semantic information among different components in a unified network. Given a remote sensing image as input, the HSP framework can detect instances of semantic objects belonging to certain categories in an end-to-end way. First, the multiscale representation is captured by a basic feature pyramid network, which can hierarchically combine spatial attention details and the global semantic structure in order to learn more discriminative visual features. Second, the soft-segmentation prediction is used as an auxiliary objective in the intermediate layer of our HSP; its output instance-aware semantic information can be propagated to suppress noisy background information and thereby guide the proposal generation in the region proposal network. By further propagating this hierarchical semantic information into the region of interest module, we can then predict the object category information and the corresponding horizontal and oriented bounding boxes. Comprehensive evaluations on three benchmark data sets demonstrate the superiority of our HSP to the existing state-of-the-art methods for object detection in remote sensing imagery.

Abstracts:This article investigates how to select the optimal Moderate-Resolution Imaging Spectroradiometer (MODIS) and Landsat 8 OLI image pairs for MODIS fractional snow cover (FSC) mapping using an artificial neural network (ANN). Four issues are discussed, including date selection, location selection, priority of date and location, and global and regional monitoring of MODIS FSC with ANNs. We propose using the histogram quadratic distance to define the similarity between the ANN training and the target test scene, which was used to quantify the representativeness of the training samples. We use the case study of MODIS FSC mapping of North Xinjiang, China, in the 2014&#x2013;2015 snow season as an example. Thirty-eight experiments were designed. The experimental results demonstrate that the ANN-based FSC estimation accuracy outperformed the MODIS FSC product, with an average RMSE of 0.17, <inline-formula> <tex-math notation="LaTeX">${R}$ </tex-math></inline-formula> exceeds 0.8, and the total snow cover area was estimated more accurately in most cases. For a target test scene, we preliminarily inferred that the best method is to develop an ANN using image pairs of another location with the highest similarity in the same acquisition time, using historical image pairs of the target scene with the highest similarity is the second choice, and using historical image pairs from another location with a high similarity is the third choice. For global- and regional-scale MODIS FSC mapping with ANNs, we formulated the strategy of initially determining a reasonable location and subsequently selecting the acquisition date of the image pairs to guarantee that the training data set represents the entire study area well.

Abstracts:Hyperspectral image (HSI) super-resolution (SR) is a challenging task due to the problems of texture blur and spectral distortion when the upscaling factor is large. To meet these two challenges, band attention through the adversarial learning method is proposed in this article. First, we put the SR process in a generative adversarial network (GAN) framework, so that the resulted high-resolution HSI can keep more texture details. Second, different from the other band-by-band SR method, the input of our method is of full bands. In order to explore the correlation of spectral bands and avoid the spectral distortion, a band attention mechanism is proposed in our generative network. A series of spatial&#x2013;spectral constraints or loss functions is imposed to guide the training of our generative network so as to further alleviate spectral distortion and texture blur. The experiments on the Pavia and Cave data sets demonstrate that the proposed GAN-based SR method can yield very high-quality results, even under large upscaling factor (e.g., <inline-formula> <tex-math notation="LaTeX">$8times $ </tex-math></inline-formula>). More importantly, it can outperform the other state-of-the-art methods by a margin which demonstrates its superiority and effectiveness.

Abstracts:In this article, we evaluate the impact of assimilating spaceborne synthetic aperture radar (SAR) data in an Arctic regional ice analysis system over a year cycle. Ice and water information was automatically extracted from more than 7000 RADARSAT-2 HH-HV ScanSAR Wide images acquired over the Canadian Arctic and adjacent waters throughout the entire year 2013. A quality-control procedure was specifically developed and applied to reduce the number of erroneous SAR retrievals. To assess the impact of SAR ice and water retrievals on the Environment and Climate Change Canada (ECCC) Regional Ice-Ocean Prediction System (RIOPS) ice concentration analyses, we designed a set of data assimilation experiments with and without the inclusion of SAR retrievals. Our verification results suggest that the assimilation of SAR-derived retrievals considerably improves ice concentration analyses in the situations where high spatial resolution is important (e.g., near land and over small inland lakes). Furthermore, SAR retrievals are particularly useful over the areas where the Canadian Ice Service&#x2019;s (CIS) manually derived ice products (such as Image Analyses, daily and weekly ice charts) are not available or have limited coverage. The three-satellite RADARSAT Constellation Mission (RCM) launched in June 2019 will significantly increase the temporal frequency of SAR data. According to the most recent CIS estimate, more than 54 000 RCM images a year will be acquired over the CIS areas of interest. Therefore, the assimilation of SAR retrievals from RCM should further enhance automated ice concentration analyses products.

Abstracts:The quantity and quality of training samples have a great influence on the performance of most hyperspectral image classification approaches. However, in a real scenario, manually annotating a large number of accurate training samples is extremely labor-intensive and time-consuming. In this article, a multilabel training sample augmentation method is proposed. Instead of giving an exact label to each pixel, we just precisely label a small number of pixels by giving them a single label (called single-label samples) and annotate a large number of pixels in certain regions together by giving them multiple labels (called multilabel samples). Furthermore, in order to make full use of the multilabel training samples, a superpixel segmentation and recursive filtering-based method is proposed. The proposed method consists of the following major steps: recursive filtering-based feature extraction, superpixel-based segmentation, and spectral&#x2013;spatial similarity-based mislabeled sample removal. Experimental results demonstrate that the proposed method can significantly improve the classification accuracy of multiple classifiers by using the multilabel training samples.