Gaofen-3 Synthetic Aperture Radar Images Research Articles

Validation of Surface Waves Investigation and Monitoring Data against Simulation by Simulating Waves Nearshore and Wave Retrieval from Gaofen-3 Synthetic Aperture Radar Image

The Chinese-French Oceanography SATellite (CFOSAT) jointly developed by the Chinese National Space Agency (CNSA) and the Centre National d’Etudes Spatiales (CNES) of France carries a wave spectrometer (Surface Waves Investigation and Monitoring, SWIM). SWIM has one nadir and five off-nadir beams to measure ocean surface waves. These near-nadir beams range from 0° to 10° at an interval of 2°. In this work, we investigated the performance of wave parameters derived from wave spectra measured by SWIM at off-nadir beams during the period 2020 to December 2022, e.g., incidence angles of 6°, 8° and 10°, which were collocated with the wave simulated by Simulating Waves Nearshore (SWAN). The validation of SWAN-simulated significant wave heights (SWHs) against National Data Buoy Center (NDBC) buoys of National Oceanic and Atmospheric Administration (NOAA) exhibited a 0.42 m root mean square error (RMSE) in the SWH. Our results revealed a RMSE of 1.02 m for the SWIM-measured SWH in the East Pacific Ocean compared with the SWH simulated by SWAN, as well as a 0.79 correlation coefficient (Cor) and a 1.17 squared error (Err) for the wave spectrum at an incidence angle of 10°, which are better than those (i.e., the RMSEs were > 1.1 m with Cors < 0.76 and Errs > 1.2) achieved at other incidence angles of SWH up to 14 m. This analysis indicates that the SWIM product is a relevant resource for wave monitoring over global seas. The collocated wave retrievals for more than 300 cases from Gaofen-3 (GF-3) synthetic aperture radar (SAR) images in China Seas were also used to verify the accuracy of SWIM-measured wave spectra. The energy of the SWIM-measured wave spectra represented by SWH was found to decrease with an increasing incidence angle in a case study. Moreover, the SWIM-measured wave spectra were most consistent with the SAR-derived wave spectra at an incidence angle of 10°, yielding a 0.77 Cor and 1.98 Err between SAR-derived and SWIM wave spectra under regular sea state conditions (SWH < 2 m). The error analysis indicates that the difference in SWH between SWIM at an incidence angle of 10° and SWAN has an increasing tendency with the growth in sea surface wind and sea state and it stabilizes to be 0.6 m at SWH > 4 m; however, the current and sea level have less influence on the uncertainties of the SWIM product.

Revealing the Land Subsidence Deceleration in Beijing (China) by Gaofen-3 Time Series Interferometry

Revealing the land subsidence in Beijing, China, induced by the massive groundwater extraction in the past three decades, is important to mitigate the hazards and protect the residences and infrastructure. Many SAR (Synthetic Aperture Radar) datasets have been successfully applied to reveal the land subsidence over Beijing in previous research, while few works were achieved on land subsidence revealed by time-series InSAR (Interferometric Synthetic Aperture Radar) with Gaofen-3 SAR images. In this study, we successfully perform the time-series InSAR analysis with Gaofen-3 SAR images to extract the land subsidence in Beijing from 2020 to 2021. The Sentinel-1 SAR images were used to assess the accuracy of Gaofen-3 images. The subsidence scale and extent are consistent in detected major subsidence bowls between the two datasets. The spatial–temporal evolution and the deceleration of Beijing land subsidence were revealed by comparing with the Sentinel-1 results from 2017 to 2020. Moreover, we evaluated the interferometric performance of Gaofen-3 satellite SAR imagery and analyzed the main factors that mostly influence the coherence and quality of interferograms. Our results proved that the long perpendicular baselines decrease the coherence seriously over the study area, and the artifacts induced by inaccurate orbit information reduce the quality of the Gaofen-3 interferograms. Refining and removing the two main artifacts could improve the quality of interferograms formed by Gaofen-3 SAR images.

Fast and Accurate Refocusing for Moving Ships in SAR Imagery Based on FrFT

Synthetic aperture radar (SAR) is capable of monitoring the ocean all day, regardless of weather conditions. However, moving ships exhibit azimuth defocus in SAR images, which severely hampers ship recognition performance. Ships typically move in a linear motion at sea. For refocusing linear moving ships, existing SAR autofocus algorithms cannot accurately extract defocus information and require multiple iterations. To overcome the poor focusing quality and high computational complexity of existing refocusing algorithms, this paper proposes a fast and accurate refocusing scheme for moving ships in SAR imagery based on Fractional Fourier Transform (FrFT). Firstly, the azimuth line with the strongest energy in the SAR image is selected as the best azimuth line representing its motion property. Then, according to the entropy variation law of the azimuth line after FrFT, the azimuth line’s optimal rotation order is determined by the proposed minimum entropy search method, which can accurately and quickly obtain defocus information. In the final refocusing module, the scheme provides two ways, i.e., fast or fine refocusing approaches, to generate well-focused images. The fast refocusing approach performs FrFT on each azimuth line at the optimal rotation order of the best azimuth line. The fine refocusing approach takes the optimal rotation order of the best azimuth line as the initial value and further searches for the optimal rotation order of other azimuth lines. In order to verify the effectiveness of the proposed scheme, experiments are carried out on a number of Gaofen-3 SAR images in different acquisition modes. The experimental results show that the proposed fast refocusing approach can achieve the fastest speed, which is 2.1% of the traditional FrFT-based method’s processing time. Moreover, the proposed fine refocusing approach has the best focusing performance, achieving the lowest image entropy among existing methods.

Mapping Growing Stem Volume Using Dual-Polarization GaoFen-3 SAR Images in Evergreen Coniferous Forests

Unaffected by cloud cover and solar illumination, synthetic aperture radar (SAR) images have great capability to map forest growing stem volume (GSV) in complex biophysical environments. Up to now, c-band dual-polarization Gaofen-3 (GF-3) SAR images, acquired by the first Chinese civilian satellite equipped with multi-polarized modes, are rarely applied in mapping forest GSV. To evaluate the capability of dual-polarization GF-3 SAR images in mapping forest GSV, several proposed derived features were initially extracted by mathematical operations and applied to obtain optimal feature sets by different feature sorting methods and feature selection methods. Then, the maps of GSV in an evergreen coniferous forest were inverted by various machine learning algorithms and stacking ensemble learning methods with different strategies. The results implied that backscattering coefficients and partially proposed derived features showed high sensitivity to the forest GSV, and the saturation phenomenon also obviously occurred once the forest GSV was larger than 300 m3/ha. Furthermore, the results showed that the accuracy of the mapped GSV was significantly improved using the stacking ensemble learning methods. Using various optimal feature sets and base models (MLR, KNN, SVM, and RF), the rRMSE values mainly ranged from 30% to 40%. After using the stacking ensemble learning methods, the values of rRMSE ranged from 16.71% to 20.51%. This confirmed that dual-polarization GF-3 images have great potential to map forest GSV in evergreen coniferous forests.

Towards real-time detection of ships and wakes with lightweight deep learning model in Gaofen-3 SAR images

Synthetic aperture radar (SAR) is an important tool for achieving accurate and efficient monitoring of maritime traffic by identifying ships and wakes. Ship wake can not only assist in ship detection, but also invert ship navigation information, which has attracted wide attention. However, most of the work on ship wake detection is still through non-deep learning methods. Limited by factors such as unclear features, complex background interference and insufficient label information in large-scale SAR images, deep learning methods have developed slowly in the field of ship wake detection. In this paper, we propose a lightweight deep learning network for ship and wake detection. In the framework, a YOLO structure based deep learning is implemented to achieve ship and wake classification and localization. For this purpose, we initially create a Gaofen-3 SAR dataset containing about 1000 pairs of ships and wakes. To streamline the deep learning network, based on the background and pixel characteristics of SAR images, we design a combined convolutional module in combination with an attention mechanism, which replaces the convolutional layers in the backbone to extract more diverse feature maps at a low cost. In addition, in view of the geometric features of ship wake, the optimized design is conducted in the target prediction stage, and the angle-related loss function is developed to guide the training of the network, which not only ensures the detection of the ship target, but also greatly improves the accuracy of wake line detection. Experimental results show that the developed lightweight network achieves accurate and stable detection of ships and wakes in the dataset, and it has excellent generalization ability in complex background and weak target detection. The performance is comparable to the state-of-the-art models, where the average precision of ship and wake detection reaches 77.86% and 97.29%, respectively. Particularly, the average angle error between the predicted and actual values of wake lines is reduced by 3.476°. Moreover, the low energy consumption (model size is one-sixth of YOLOv4) and real-time detection (>23.1 FPS) of the algorithm offer great potential for embedded deployment.

Preliminary analysis of wave retrieval from Chinese Gaofen-3 SAR imagery in the Arctic Ocean

ABSTRACT Synthetic aperture radar (SAR) is a powerful tool for sea surface monitoring. The 2017–2020 mission of the Chinese Gaofen-3 (GF-3) over the Arctic Ocean provided a unique opportunity to investigate the feasibility of wave retrieval in the Arctic Ocean. In this study, several GF-3 SAR images acquired in quad-polarization (QPS) mode were applied for wave retrieval using two theoretical-based algorithms, i.e. the parameterized first-guess spectrum method (PFSM) (with improved tilt modulation) and quad-polarization (Q-P). These images were collocated with wave fields simulated using the third-generation numeric wave model WAVEWATCH-III (WW3) and measurements from the Haiyang-2B (HY-2B) altimeter. A comparison of the HY-2B winds with the wind speeds generated by the geophysical model function (GMF) CMOD5.N showed a 3.36 m/s root mean square error (RMSE) with a correlation (COR) of 0.66. The retrieved significant wave height (SWH) validated against the WW3-simulated results had a 0.50 m RMSE using PFSM and 0.91 m RMSE with the Q-P algorithm. These results show that PFSM is suitable for wave retrieval from GF-3 SAR images. We also used PFSM to investigate the characteristics of the wave spectrum in the presence of sea ice. Although the analysis concludes that GF-3 SAR has the capability for wave monitoring in Arctic Ocean due to the high spatial resolution of SAR-derived wave spectra, an optimal wave retrieval algorithm needs to be developed for improving the retrieval accuracy.

Evaluation of wave retrieval for Chinese Gaofen-3 synthetic aperture radar

ABSTRACT The goal of this study was to investigate the performance of a spectral-transformation wave retrieval algorithm and confirm the accuracy of wave retrieval from C-band Chinese Gaofen-3 (GF-3) Synthetic Aperture Radar (SAR) images. More than 200 GF-3 SAR images of the coastal China Sea and the Japan Sea for dates from January to July 2020 were acquired in the Quad-Polarization Strip (QPS) mode. The images had a swath of 30 km and a spatial resolution of 8 m pixel size. They were processed to retrieve Significant Wave Height (SWH), which is simulated from a numerical wave model called Simulating WAves Nearshore (SWAN). The first-guess spectrum is essential to the accuracy of Synthetic Aperture Radar (SAR) wave spectrum retrieval. Therefore, we proposed a wave retrieval scheme combining the theocratic-based Max Planck Institute Algorithm (MPI), a Semi-Parametric Retrieval Algorithm (SPRA), and the Parameterized First-guess Spectrum Method (PFSM), in which a full wave-number spectrum and a non-empirical ocean spectrum proposed by Elfouhaily are applied. The PFSM can be driven using the wind speed without calculating the dominant wave phase speed. Wind speeds were retrieved using a Vertical-Vertical (VV) polarized geophysical model function C-SARMOD2. The proposed algorithm was implemented for all collected SAR images. A comparison of SAR-derived wind speeds with European Center for Medium-Range Weather Forecasts (ECMWF) ERA-5 data showed a 1.95 m/s Root-Mean-Squared Error (RMSE). The comparison of retrieved SWH with SWAN-simulated results demonstrated a 0.47 m RMSE, which is less than the 0.68 m RMSE of SWH when using the PFSM algorithm.

Radiometric Principle-Based Radiometric Normalization Method for SAR Images Mosaic

Radiometric normalization minimizes the radiometric inconsistencies between images in synthetic aperture radar (SAR) images mosaic. However, the radiometric principle of SAR have not been fully considered by the existing methods. To this issue, a radiometric principle-based radiometric normalization method is proposed. First, image areas with consistent coverage areas in object space and approximate rough ground surface between images are extracted as adjustment candidates based on imaging principle. Then, considering the radiometric characteristic of SAR image, all images are taken as a whole to solve the radiometric adjustment model, which transfers radiometric normalization into the least square optimization. Finally, a global quantization strategy is used to ensure radiometric consistency during orthorectification and mosaicking. The experimental results of Gaofen-3 SAR images demonstrate that the proposed method has the best performance, and can effectively and stably eliminate the radiometric differences between images.

Comparing Methods for Segmenting Supra-Glacial Lakes and Surface Features in the Mount Everest Region of the Himalayas Using Chinese GaoFen-3 SAR Images

Glaciers and numerous glacial lakes that are produced by glacier melting are key indicators of climate change. Often overlooked, supra-glacial lakes develop in the melting area in the low-lying part of a glacier and appear to be highly variable in their size, shape, and location. The lifespan of these lakes is thought to be quite transient, since the lakes may be completely filled by water and burst out within several weeks. Changes in supra-glacial lake outlines and other surface features such as supra-glacial rivers and crevasses on the glaciers are useful indicators for the direct monitoring of glacier changes. Synthetic aperture radar (SAR) is not affected by weather and climate, and is an effective tool for study of glaciated areas. The development of the Chinese GaoFen-3 (GF-3) SAR, which has high spatial and temporal resolution and high-precision observation performance, has made it possible to obtain dynamic information about glaciers in more detail. In this paper, the classical Canny operator, the variational B-spline level-set method, and U-Net-based deep-learning model were applied and compared to extract glacial lake outlines and other surface features using different modes and Chinese GF-3 SAR imagery in the Mount Everest Region of the Himalayas. Particularly, the U-Net-based deep-learning method, which was independent of auxiliary data and had a high degree of automation, was used for the first time in this context. The experimental results showed that the U-Net-based deep-learning model worked best in the segmentation of supra-glacial lakes in terms of accuracy (Precision = 98.45% and Recall = 95.82%) and segmentation efficiency, and was good at detecting small, elongated, and ice-covered supra-glacial lakes. We also found that it was useful for accurately identifying the location of supra-glacial streams and ice crevasses on glaciers, and quantifying their width. Finally, based on the time series of the mapping results, the spatial characteristics and temporal evolution of these features over the glaciers were comprehensively analyzed. Overall, this study presents a novel approach to improve the detection accuracy of glacier elements that could be leveraged for dynamic monitoring in future research.

Ship Detection in SAR Images by Aggregating Densities of Fisher Vectors: Extension to a Global Perspective

Fisher vectors (FVs) can capture multiple order information from superpixels (SPs) in synthetic aperture radar (SAR) images. Existing FV-based ship detectors mainly exploit the local contrast of FVs (LCFVs) but do not consider their global density features. This may lead to degraded performance in terms of discrimination between ship targets and the complex sea clutter. In this article, two new global cues from FVs are designed based on the fact that target FVs exhibit much lower densities than those of clutter FVs and also have large distances to the latter. Our two new global cues can suppress the sea clutter and significantly enhance ship targets throughout the SAR image. We also design an improved local cue from FVs for ship detection, in which the intensity contrast of SPs is incorporated into the existing LCFV indicator to reduce false alarms. By fusing the above two new global cues (and an improved local cue from FVs), we propose a new method for ship detection in SAR images. Experimental results based on Gaofen-3 SAR images show that the newly proposed detector provides better detection performance than other state-of-the-art detectors, especially in the presence of strong and highly heterogeneous sea clutter.

Revisit of a Case Study of Spilled Oil Slicks Caused by the Sanchi Accident (2018) in the East China Sea

Marine oil spills occur suddenly and pose a serious threat to ecosystems in coastal waters. Oil spills continuously affect the ocean environment for years. In this study, the oil spill caused by the accident of the Sanchi ship (2018) in the East China Sea was hindcast simulated using the oil particle-tracing method. Sea-surface winds from the European Centre for Medium-Range Weather Forecasts (ECMWF), currents simulated from the Finite-Volume Community Ocean Model (FVCOM), and waves simulated from the Simulating WAves Nearshore (SWAN) were employed as background marine dynamics fields. In particular, the oil spill simulation was compared with the detection from Chinese Gaofen-3 (GF-3) synthetic aperture radar (SAR) images. The validation of the SWAN-simulated significant wave height (SWH) against measurements from the Jason-2 altimeter showed a 0.58 m root mean square error (RMSE) with a 0.93 correlation (COR). Further, the sea-surface current was compared with that from the National Centers for Environmental Prediction (NCEP) Climate Forecast System Version 2 (CFSv2), yielding a 0.08 m/s RMSE and a 0.71 COR. Under these circumstances, we think the model-simulated sea-surface currents and waves are reliable for this work. A hindcast simulation of the tracks of oil slicks spilled from the Sanchi shipwreck was conducted during the period of 14–17 January 2018. It was found that the general track of the simulated oil slicks was consistent with the observations from the collected GF-3 SAR images. However, the details from the GF-3 SAR images were more obvious. The spatial coverage of oil slicks between the SAR-detected and simulated results was about 1 km2. In summary, we conclude that combining numerical simulation and SAR remote sensing is a promising technique for real-time oil spill monitoring and the prediction of oil spreading.

Ship Detection in SAR Images via Enhanced Nonnegative Sparse Locality-Representation of Fisher Vectors

As a powerful coding strategy for superpixels in synthetic aperture radar (SAR) images, Fisher vector (FV) lies in a low-dimensional subspace and can be sparsely represented as a linear combination of training samples. The existing ship detection methods based on FVs often consider the Euclidean distances between target FVs and clutter FVs, where the subspace features of FVs are generally not exploited. In this article, we propose a new ship detection algorithm based on nonnegative sparse locality-representation (NSLR) to exploit the subspace features of FVs. The proposed NSLR method is based on the assumption that FVs of superpixels in SAR images are sparsely represented by the dictionary of background sea clutter only under a null hypothesis. In addition, we propose two FV-based filters to enhance the robustness of our newly developed NSLR to heterogeneous sea clutter environments by further exploiting the intrinsic features of ship targets in terms of intensity and spatiality. The experimental results based on Gaofen-3 SAR images demonstrate that the proposed NSLR detection method provides higher target-to-clutter contrast and achieves better detection performance than other commonly used ship detection algorithms.

Automatic Extraction of Green Tide Using Dual Polarization Chinese GF-3 SAR Images

Yu, H.F.; Wang, C.Y.; Sui, Y.; Li, J.H., and Chu, J.L., 2020. Automatic extraction of green tide using dual polarization China GF-3 SAR images. In: Jung, H.-S.; Lee, S.; Ryu, J.-H., and Cui, T.W. (eds.), Advances in Geospatial Research of Coastal Environments. Journal of Coastal Research, Special Issue No. 102, pp. 318-325. Coconut Creek (Florida), ISSN 0749-0208.Traditional optical images are not suitable for the all-weather observation of green tide due to the fact that they are greatly affected by clouds. Synthetic aperture radar (SAR) on the other hand, is able to realize all-day and all-weather observation due to a lower sensitivity to clouds, rain and fog. SAR images are therefore an effective supplementary data for the monitoring of green tide. Due to the differences in brightness and noise across the different areas in an image, it is difficult to use the same threshold in order to extract all the green tide information from the image. Based on the iterative threshold method and the histogram bimodal method, this paper presents a new method for the automatic detection of green tide that uses adaptive thresholds for Gaofen-3 (GF-3) satellite dual polarization SAR remote sensing images. In this study, firstly, a sliding window is used to segment the image into multiple sub-images of the same size; secondly, the iterative threshold method is used to obtain the green tide and seawater samples that have the theoretical bimodal structure from each sub-image; then, using the histogram bimodal method, the detection threshold is calculated automatically; and finally, using the threshold segmentation, the green tide areas are extracted from the images. In order to verify the proposed method, the multi-scale segmentation method and the proposed method are both used to detect green tide in China Yellow Sea. The study results show that in the case of GF-3 SAR FSII images, the proposed automatic detection method is superior to the multi-scale segmentation method, as it not only improves the accuracy of green tide detection, but also realizes the automation of green tide extraction. Furthermore, Cross-polarization (HV) images may be more suitable than co-polarization (HH) images for the extraction of green tide due to their lower noise level.

Co-seismic deformation of the 2017 M s 7.0 Jiuzhaigou Earthquake observed with GaoFen-3 interferometry

ABSTRACT The first application of GaoFen-3 synthetic aperture radar (SAR) interferometry to monitoring co-seismic deformation of an earthquake is presented in this paper. First, two GaoFen-3 SAR images spanning the 2017 surface-wave magnitude () 7.0 Jiuzhaigou earthquake were mosaicked, coregistered and processed into an interferogram with visible fringes. After the removal of severe orbital errors and unwrapping, the co-seismic deformation field is obtained. Finally, the fault parameters and slip distribution are determined based on the deformation field. To evaluate the performance of GaoFen-3 in this case, the interferogram, co-seismic deformation field, derived fault parameters and slip distribution are compared with those derived from Radarsat-2, Sentinel-1A and Advanced Land Observing Satellite-2 (ALOS-2) Phased Array type L-band Synthetic Aperture Radar-2 (PALSAR-2). The comparisons between the deformation field obtained from GaoFen-3 and those obtained from Radarsat-2 and Sentinel-1A show Pearson’s correlation coefficient () values 90%, demonstrating that GaoFen-3 interferometry can measure the co-seismic deformation effectively. The resultant slip model obtained by GaoFen-3 interferometric measurements shows a simple, elliptical pattern of slip, reaching a maximum of approximately 1 m at depths of 10–13 km. The rake angle is almost , which suggests left-lateral strike-slip motion, consistent with previous studies.

Water Body Detection in High-Resolution SAR Images With Cascaded Fully-Convolutional Network and Variable Focal Loss

The water body detection in high-resolution synthetic aperture radar (SAR) images is a challenging task due to the changing interference caused by multiple imaging conditions and complex land backgrounds. Inspired by the excellent adaptability of deep neural networks (DNNs) and the structured modeling capabilities of probabilistic graphical models, the cascaded fully-convolutional network (CFCN) is proposed to improve the performance of water body detection in high-resolution SAR images. First, for the resolution loss caused by convolutions with large stride in traditional convolutional neural network (CNN), the fully-convolutional upsampling pyramid networks (UPNs) are proposed to suppress this loss and realize pixel-wise water body detection. Then considering blurred water boundary, the fully-convolutional conditional random fields (FC-CRFs) are introduced to UPNs, which reduce computational complexity and lead to the automatic learning of Gaussian kernels in CRFs and the higher boundary accuracy. Furthermore, to eliminate the inefficient training caused by imbalanced categorical distribution in the training data set, a novel variable focal loss (VFL) function is proposed, which replaces the constant weighting factor of focal loss with the frequency-dependent factor. The proposed methods can not only improve the pixel accuracy and boundary accuracy but also perform well in detection robustness and speed. Results of GaoFen-3 SAR images are presented to validate the proposed approaches.

A New Framework for Automatic Airports Extraction from SAR Images Using Multi-Level Dual Attention Mechanism

The detection of airports from Synthetic Aperture Radar (SAR) images is of great significance in various research fields. However, it is challenging to distinguish the airport from surrounding objects in SAR images. In this paper, a new framework, multi-level and densely dual attention (MDDA) network is proposed to extract airport runway areas (runways, taxiways, and parking lots) in SAR images to achieve automatic airport detection. The framework consists of three parts: down-sampling of original SAR images, MDDA network for feature extraction and classification, and up-sampling of airports extraction results. First, down-sampling is employed to obtain a medium-resolution SAR image from the high-resolution SAR images to ensure the samples (500 × 500) can contain adequate information about airports. The dataset is then input to the MDDA network, which contains an encoder and a decoder. The encoder uses ResNet_101 to extract four-level features with different resolutions, and the decoder performs fusion and further feature extraction on these features. The decoder integrates the chained residual pooling network (CRP_Net) and the dual attention fusion and extraction (DAFE) module. The CRP_Net module mainly uses chained residual pooling and multi-feature fusion to extract advanced semantic features. In the DAFE module, position attention module (PAM) and channel attention mechanism (CAM) are combined with weighted filtering. The entire decoding network is constructed in a densely connected manner to enhance the gradient transmission among features and take full advantage of them. Finally, the airport results extracted by the decoding network were up-sampled by bilinear interpolation to accomplish airport extraction from high-resolution SAR images. To verify the proposed framework, experiments were performed using Gaofen-3 SAR images with 1 m resolution, and three different airports were selected for accuracy evaluation. The results showed that the mean pixels accuracy (MPA) and mean intersection over union (MIoU) of the MDDA network was 0.98 and 0.97, respectively, which is much higher than RefineNet and DeepLabV3. Therefore, MDDA can achieve automatic airport extraction from high-resolution SAR images with satisfying accuracy.

An Adaptive GaoFen-3 SAR Wind Field Retrieval Algorithm Based on Information Entropy

Chinese GaoFen-3 (GF-3) synthetic aperture radar (SAR) imagery acquired in standard strip mode can capture streak features caused by wind over the sea surface, which are commonly known as wind streaks. To retrieve the sea surface wind field from a GF-3 SAR image, an adaptive wind retrieval algorithm based on information entropy is proposed in this paper. This algorithm can adaptively extract the optimal wavenumber range of each subimage based on maximum information entropy. After all subimages are processed for possible wind streak direction solutions, a circle median filter is used to determine the unique wind streak direction solution for each subimage. Twelve GF-3 standard strip SAR images are inverted to validate the proposed algorithm. The experiments show that when the GF-3 image data are compared to the European Centre for Medium-Range Weather Forecasts reanalysis wind field data, the root mean square error of wind speed and direction retrieved by the algorithm are 1.29 m/s and 15.33°, respectively, and when they are compared to the National Data Buoy Center buoy wind data, the mean absolute deviation errors of wind speed and direction are 0.57 m/s and 9.95°, respectively. These comparisons demonstrate that the proposed algorithm can effectively retrieve wind field information from GF-3 SAR images.

Geospatial Contextual Attention Mechanism for Automatic and Fast Airport Detection in SAR Imagery

The automatic extraction of airport runway areas from high-resolution Synthetic Aperture Radar (SAR) images is of great research significance in the military and civilian fields. However, it is still challenging to distinguish the airport from surrounding objects in SAR images. In this article, a new framework is proposed to extract airport runway areas (runways, taxiways, packing lots, and aircrafts) in a fast and automatic manner. The framework is based on the Geospatial Contextual Attention Mechanism (GCAM) for geospatial feature learning and classification, which is employed together with the down-sampling and coordinate mapping modules. To evaluate the performance of the proposed framework, three large-scale Gaofen-3 SAR images with 1m resolution are utilized in the experiment. According to the results, Mean Pixels Accuracy (MPA) and Mean Intersection Over Union (MIOU) of the GCAM are 0.9850 and 0.9536, respectively, which outperform RefineNet, DeepLabV3+, and MDDA. The training time of GCAM for the dataset is 2.25h, and the average testing time for the five SAR images is only 18.15s. Therefore, GCAM can offer rapid and automatic airport detection from high-resolution SAR images with high accuracy, which can further be employed to mark the airport to greatly improve the detection accuracy of the aircrafts.

Random cross-observation intensity consistency method for large-scale SAR images mosaics: An example of Gaofen-3 SAR images covering China

Large-scale microwave remote sensing research requires multiple synthetic aperture radar (SAR) image mosaics. However, owing to multiple factors, different SAR images frequently exhibit intensity differences. This leads to poor intensity continuity after mosaicking, making SAR image interpretation extremely difficult. The existing method assumes that SAR images have been subject to strict radiometric calibration and mainly focuses on intensity balancing to minimize the intensity difference between adjacent images. This method is not effective when the intensity difference between SAR images is large. In this paper, a novel random cross-observation intensity consistency (RCOIC) algorithm is proposed to eliminate the intensity differences for SAR image mosaics. The proposed algorithm is based on global intensity constraints. First, source images are downsampled. Then, the gain correction coefficients of each image in a study area are accurately estimated through random cross-observation. Further, multi-view local intensity correction of each image in the study area is performed using the overlapping areas of adjacent images, and a low-resolution intensity reference map covering the study area is generated. Finally, the intensities of high-resolution source images are corrected by constructing local independent linear models of images based on the low-resolution intensity reference map. Gaofen-3 satellite images with large ranges and large time differences, covering most of China's land area, were selected to test the effect of the newly developed algorithm. After processing, the overall visual effect was good after mosaicking, the intensity transitions at the edges of the images were smooth, the intensity distributions corresponded better to actual ground objects, and the statistical result of adjacent images overlapping regions was good. This validates the reliability of the algorithm.

Demonstration of Time-Series InSAR Processing in Beijing Using a Small Stack of Gaofen-3 Differential Interferograms

More and more synthetic aperture radar (SAR) satellites in orbit provide abundant data for remote sensing applications. In August 2016, China launched a new Earth observation SAR satellite, Gaofen-3 (GF-3). In this paper, we utilize a small stack of GF-3 differential interferograms to map land subsidence in Beijing (China) using the time-series SAR interferometry (InSAR) technique. The small stack of differential interferograms is generated with 5 GF-3 SAR images from March 2017 to January 2018. Orbit errors are carefully addressed and removed during differential InSAR (DInSAR) processing. Truncated singular-value decomposition (TSVD) is applied to strengthen the robustness of deformation rate estimation. To validate the results of GF-3 data, an additional deformation measurement using 26 Sentinel-1B images from March 2017 to February 2018 is carried out using the persistent scatterer interferometry (PSI) technique. By implementing a cross-comparison, we find that the retrieved results from GF-3 images and Sentinel-1 images are spatially consistent. The standard deviation of vertical deformation rate differences between two data stacks is 11.24 mm/y in the study area. The results shown in this paper demonstrate the reasonable potential of GF-3 SAR images to monitor land subsidence.