This paper describes the C-band multi-polarization Synthetic Aperture Radar (SAR) sensor for the Gaofen-3 (GF-3) mission. Based on the requirement analysis, the design of working modes and SAR payload are given. An accurate antenna model is introduced for the pattern optimization and SAR performance calculation. The paper concludes with an overview of predicted performance which was verified by in-orbit tests.

The GaoFen-3 (GF-3) satellite is the first polarimetric synthetic aperture radar (PolSAR) satellite in China. With a designed in-orbit life of 8 years, it will provide large amounts of PolSAR data for ocean monitoring, disaster reduction, and many other applications. The polarimetric data quality is essential for all these applications, so the analysis and calibration of the polarimetric error sources are very important for GF-3. In this study, we established a full-link error model for GF-3 PolSAR system. Based on this model, we comprehensively analyzed the quantitative effects of the main error sources including the composition, figured out characteristics of the phase imbalance introduced by the antenna, and pointed out the error sources which have to be corrected. Furthermore, the polarimetric correction method for GF-3 PolSAR system is proposed. Finally, assisted by several external calibration experiments, polarimetric errors of GF-3 data are efficiently corrected during in-orbit-test phase.

Synthetic aperture radar (SAR) tomography (TomoSAR) has been widely used in the three-dimensional (3D) reconstruction of urban areas using the multi-baseline (MB) SAR data. For urban scenarios, the MB SAR data are often acquired by repeat-pass using the spaceborne SAR system. Such a data stack generally has long time baselines, which result in different atmospheric disturbances of the data acquired by different tracks. These factors can lead to the presence of phase errors (PEs). PEs are multiplicative noise for observation data, which can cause diffusion and defocus in TomoSAR imaging and seriously affect the extraction of target 3D information. In this paper, we combine the methods of the block-building network (BBN) and phase gradient autofocus (PGA) to propose a novel phase compensation method called BBN-PGA. The BBN-PGA method can effectively and efficiently compensate for PEs of the MB SAR data over a wide area and improve the accuracy of 3D reconstruction of urban areas. The applicability of this proposed BBN-PGA method is proved by using simulated data and the spaceborne MB SAR data acquired by the TerraSAR-X satellite over an area in Barcelona, Spain.

Spotlight synthetic aperture radar (SAR) is a proven technique, which can provide high-resolution images as compared to those produced by traditional stripmap SAR. This paper addresses a high-resolution SAR focusing experiment based on Gaofen-3 satellite (GF-3) staring data with about 55 cm azimuth resolution and 240 MHz range bandwidth. In staring spotlight (ST) mode, the antenna always illuminates the same scene on the ground, which can extend the synthetic aperture. Based on a two-step processing algorithm, some special aspects such as curved-orbit model error correction, stop-and-go correction, and antenna pattern demodulation must be considered in image focusing. We provide detailed descriptions of all these aspects and put forward corresponding solutions. Using these suggested methods directly in an imaging module without any modification for other data processing software can make the most of the existing ground data processor. Finally, actual data acquired in GF-3 ST mode is used to validate these methodologies, and a well-focused, high-resolution image is obtained as a result of this focusing experiment.

There are two useful methods of current measurement based on synthetic aperture radar (SAR): one is along-track interferometry (ATI), and the other is Doppler centroid analysis (DCA). For the ATI method, the interferometric phase must be accurate enough for ocean current measurements. Therefore, the space-varying of phase imbalances along the range, caused by antenna phase center position error, attitude error, antenna electronic miss pointing, antenna pattern mismatch, and other reasons, cannot be ignored. Firstly, this paper mainly analyzes the above possible factors by using real GF-3 ATI data and error model simulation results. Secondly, the ocean current has been preliminarily measured by the ATI method and the DCA method, using CDOP model, based on the GF-3 ATI data of the ocean scene near Qingdao, China, which is up to around −1.45 m/s. The results of the two methods are in good agreement with the correlation coefficient of 0.98, the mean difference of −0.010 m/s, and the root mean squared error (RMSE) of 0.062 m/s. Moreover, by comparing with the current measured by high-frequency surface wave radar (HFSWR), the correctness of the analysis is further proved.

Superpixel segmentation is widely used in polarimetric synthetic aperture radar (PolSAR) image classification. However, the classification method using simple majority voting cannot easily handle evidence conflicts in a single superpixel. At present, there is no method to evaluate the quality of superpixel classification. To solve the above problems, this paper proposes a hybrid classification model based on superpixel entropy discrimination (SED), and constructs a two-level cascade classifier. Firstly, a light gradient boosting machine (LGBM) was used to process large-dimensional input features, and simple linear iterative clustering (SLIC) was integrated to obtain the primary classification results based on superpixels. Secondly, information entropy was introduced to evaluate the quality of superpixel classification, and a complex-valued convolutional neural network (CV-CNN) was used to reclassify the high-entropy superpixels to obtain the secondary classification results. Experiments with two measured PolSAR datasets show that the overall accuracy of both classification methods exceeded 97%. This method suppressed the evidence conflict in a single superpixel and the inaccuracy of superpixel segmentation. The test time of our proposed method was shorter than that of CV-CNN, and using only 55% of CV-CNN test data could achieve the same accuracy as using CV-CNN for the whole image.