Detailed and accurate information on the spatial variation of land cover and land use is a critical component of local ecology and environmental research. For these tasks, high spatial resolution images are required. Considering the trade-off between high spatial and high temporal resolution in remote sensing images, many learning-based models (e.g., Convolutional neural network, sparse coding, Bayesian network) have been established to improve the spatial resolution of coarse images in both the computer vision and remote sensing fields. However, data for training and testing in these learning-based methods are usually limited to a certain location and specific sensor, resulting in the limited ability to generalize the model across locations and sensors. Recently, generative adversarial nets (GANs), a new learning model from the deep learning field, show many advantages for capturing high-dimensional nonlinear features over large samples. In this study, we test whether the GAN method can improve the generalization ability across locations and sensors with some modification to accomplish the idea “training once, apply to everywhere and different sensors” for remote sensing images. This work is based on super-resolution generative adversarial nets (SRGANs), where we modify the loss function and the structure of the network of SRGANs and propose the improved SRGAN (ISRGAN), which makes model training more stable and enhances the generalization ability across locations and sensors. In the experiment, the training and testing data were collected from two sensors (Landsat 8 OLI and Chinese GF 1) from different locations (Guangdong and Xinjiang in China). For the cross-location test, the model was trained in Guangdong with the Chinese GF 1 (8 m) data to be tested with the GF 1 data in Xinjiang. For the cross-sensor test, the same model training in Guangdong with GF 1 was tested in Landsat 8 OLI images in Xinjiang. The proposed method was compared with the neighbor-embedding (NE) method, the sparse representation method (SCSR), and the SRGAN. The peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) were chosen for the quantitive assessment. The results showed that the ISRGAN is superior to the NE (PSNR: 30.999, SSIM: 0.944) and SCSR (PSNR: 29.423, SSIM: 0.876) methods, and the SRGAN (PSNR: 31.378, SSIM: 0.952), with the PSNR = 35.816 and SSIM = 0.988 in the cross-location test. A similar result was seen in the cross-sensor test. The ISRGAN had the best result (PSNR: 38.092, SSIM: 0.988) compared to the NE (PSNR: 35.000, SSIM: 0.982) and SCSR (PSNR: 33.639, SSIM: 0.965) methods, and the SRGAN (PSNR: 32.820, SSIM: 0.949). Meanwhile, we also tested the accuracy improvement for land cover classification before and after super-resolution by the ISRGAN. The results show that the accuracy of land cover classification after super-resolution was significantly improved, in particular, the impervious surface class (the road and buildings with high-resolution texture) improved by 15%.

Knowledge of the exact statistical properties of the signal plays an important role in the applications of Polarimetric Synthetic Aperture Radar (PolSAR) data. In the last three decades, a considerable research effort has been devoted to finding accurate statistical models for PolSAR data, and a number of distributions have been proposed. In order to see the differences of various models and to make a comparison among them, a survey is provided in this paper. Texture models, which could capture the non-Gaussian behavior observed in high resolution data, and yet keep a compact mathematical form, are mainly explained. Probability density functions for the single look data and the multilook data are reviewed, as well as the advantages and applicable context of those models. As a summary, challenges in the area of statistical analysis of PolSAR data are also discussed.

Detection of coastline in synthetic aperture radars (SARs) is difficult due to the presence of speckle effect and strong signal return from wind-roughened, wave-modulated sea. This paper presents a new approach to detect coastlines from SAR images by integrating watershed transformation and gradient vector flow (GVF) snake model. Several improvements have been made to improve the accuracy and efficiency of coastline detection. First, ratio of averages edge detector is used to produce gradient maps suitable for watershed transformation. Second, an improved GVF snake model is presented, which exploits two external constraint forces to make the curve evolution more controllable. We name it controllable GVF (CGVF) snake model. Third, a coarse-fine processing scheme is employed, in which watershed transformation is performed on a coarse-resolution image to obtain the initial contours for CGVF snake model, and then CGVF snake model is used to refine the roughly detected coastline at fine resolution. Experimental results on Envisat-ASAR and TerraSAR-X images show that with only a modest computational burden, the new approach produces a good match between the detected coastline and the true one.

Timely and accurate crop type mapping is a critical prerequisite for the estimation of water availability and environmental carrying capacity. This research proposed a method to integrate time series Sentinel-1 (S1) and Sentinel-2 (S2) data for crop type mapping over oasis agricultural areas through a case study in Northwest China. Previous studies using synthetic aperture radar (SAR) data alone often yield quite limited accuracy in crop type identification due to speckles. To improve the quality of SAR features, we adopted a statistically homogeneous pixel (SHP) distributed scatterer interferometry (DSI) algorithm, originally proposed in the interferometric SAR (InSAR) community for distributed scatters (DSs) extraction, to identify statistically homogeneous pixel subsets (SHPs). On the basis of this algorithm, the SAR backscatter intensity was de-speckled, and the bias of coherence was mitigated. In addition to backscatter intensity, several InSAR products were extracted for crop type classification, including the interferometric coherence, master versus slave intensity ratio, and amplitude dispersion derived from SAR data. To explore the role of red-edge wavelengths in oasis crop type discrimination, we derived 11 red-edge indices and three red-edge bands from Sentinel-2 images, together with the conventional optical features, to serve as input features for classification. To deal with the high dimension of combined SAR and optical features, an automated feature selection method, i.e., recursive feature increment, was developed to obtain the optimal combination of S1 and S2 features to achieve the highest mapping accuracy. Using a random forest classifier, a distribution map of five major crop types was produced with an overall accuracy of 83.22% and kappa coefficient of 0.77. The contribution of SAR and optical features were investigated. SAR intensity in VH polarization was proved to be most important for crop type identification among all the microwave and optical features employed in this study. Some of the InSAR products, i.e., the amplitude dispersion, master versus slave intensity ratio, and coherence, were found to be beneficial for oasis crop type mapping. It was proved the inclusion of red-edge wavelengths improved the overall accuracy (OA) of crop type mapping by 1.84% compared with only using conventional optical features. In comparison, it was demonstrated that the synergistic use of time series Sentinel-1 and Sentinel-2 data achieved the best performance in the oasis crop type discrimination.

A new method to identify short-rotation eucalyptus plantations by exploring both the changing pattern of vegetation indices due to tree crop rotation and spectral characteristics of eucalyptus in the red-edge region is presented. It can be adopted to produce eucalyptus maps of high spatial resolution (30 m) at large scales, with the use of open remote sensing images from Landsat 8 Operational Land Imager (OLI), MODerate resolution Imaging Spectroradiometer (MODIS), and Sentinel-2 MultiSpectral Instrument (MSI), as well as a free cloud computing platform, Google Earth Engine (GEE). The method is composed of three main steps. First, a time series of Enhanced Vegetation Index (EVI) is constructed from Landsat data for each pixel, and a statistical hypothesis testing is followed to determine whether the pixel belongs to a tree plantation or not based on the idea that tree crops should be harvested in a specific period. Then, a broadleaf/needleleaf classification is applied to distinguish eucalyptus from coniferous trees such as pine and fir using the red-edge bands of Sentinel-2 data. Refinements based on superpixel are performed at last to remove the salt-and-pepper effects resulted from per-pixel detection. The proposed method allows gaps in the time series that are very common in tropical and subtropical regions by employing time series segmentation and statistical hypothesis testing, and could capture forest disturbances such as conversion of natural forest or agricultural lands to eucalyptus plantations emerged in recent years by using a short observing time. The experiment in Guangxi province of China demonstrated that the method had an overall accuracy of 87.97%, with producer’s accuracy of 63.85% and user’s accuracy of 66.89% for eucalyptus plantations.

Accurate and efficient extraction of cultivated land data is of great significance for agricultural resource monitoring and national food security. Deep-learning-based classification of remote-sensing images overcomes the two difficulties of traditional learning methods (e.g., support vector machine (SVM), K-nearest neighbors (KNN), and random forest (RF)) when extracting the cultivated land: (1) the limited performance when extracting the same land-cover type with the high intra-class spectral variation, such as cultivated land with both vegetation and non-vegetation cover, and (2) the limited generalization ability for handling a large dataset to apply the model to different locations. However, the “pooling” process in most deep convolutional networks, which attempts to enlarge the sensing field of the kernel by involving the upscale process, leads to significant detail loss in the output, including the edges, gradients, and image texture details. To solve this problem, in this study we proposed a new end-to-end extraction algorithm, a high-resolution U-Net (HRU-Net), to preserve the image details by improving the skip connection structure and the loss function of the original U-Net. The proposed HRU-Net was tested in Xinjiang Province, China to extract the cultivated land from Landsat Thematic Mapper (TM) images. The result showed that the HRU-Net achieved better performance (Acc: 92.81%; kappa: 0.81; F1-score: 0.90) than the U-Net++ (Acc: 91.74%; kappa: 0.79; F1-score: 0.89), the original U-Net (Acc: 89.83%; kappa: 0.74; F1-score: 0.86), and the Random Forest model (Acc: 76.13%; kappa: 0.48; F1-score: 0.69). The robustness of the proposed model for the intra-class spectral variation and the accuracy of the edge details were also compared, and this showed that the HRU-Net obtained more accurate edge details and had less influence from the intra-class spectral variation. The model proposed in this study can be further applied to other land cover types that have more spectral diversity and require more details of extraction.