Using object-based deep learning for the urban land cover classification has become a mainstream method. This study proposed an urban land cover classification method based on segments’ object features, deep features and spatial association features. The proposed method used the synthetic semivariance function to determine the hyperparameters of the superpixel segmentation and subsequently optimized the image superpixel segmentation result. A convolutional neural network (CNN) and a graph convolutional neural network (GCN) were used to obtain segments’ deep features and spatial association features respectively. The random forest algorithm was used to classify segments based on the multi-dimension features. The results showed that the image superpixel segmentation results had the significant impact on the classification results. Compared with the pixel-based method, the segment-based methods generally yielded the higher classification accuracy. The strategy of multi-dimension feature fusion can combine the advantages of each singledimension feature to improve the classification accuracy. The proposed method utilizing multi-dimension features was more efficient than traditional methods used for the urban land cover classification. The fusion of segments’ object features, deep features and spatial association features was the best solution for achieving the urban land cover classification.

Convolutional neural networks (CNNs) are very important deep neural networks for analyzing visual imagery. However, most CNN-based methods have the problem of over-smoothing at boundaries, which is unfavorable for hyperspectral image classification. To address this problem, a spectral-spatial multiscale residual network (SSMRN) by fusing two separate deep spectral features and deep spatial features is proposed to significantly reduce over-smoothing and effectively learn the features of objects. In the implementation of the SSMRN, a multiscale residual convolutional neural network is proposed as a spatial feature extractor and a band grouping-based bi-directional gated recurrent unit is utilized as a spectral feature extractor. Considering that the importance of spectral and spatial features may vary depending on the spatial resolution of images, we combine both features with two weighting factors with different initial values that can be adaptively adjusted during the network training. To evaluate the effectiveness of the SSMRN, extensive experiments are conducted on public benchmark data sets. The proposed method can retain the detailed boundary of different objects and yield competitive results compared with several state-of-the-art methods.

The Modulation Transfer Function (MTF) is a widely used parameter to assess the quality of an imaging system. For the end user, the system MTF can be used to compare the intrinsic quality of imagery from various sources as well as analytically equalize the sharpness of multiple images from different sensors. However, due to the vibration during the satellite launch or some change in material properties, the MTF characteristic may change. As a result, imagery quality may change. However, Current researches mostly lie in how to measure MTF, and don't detailedly analyze the influence of MTF on imagery quality. Therefore, a lot of potential research work is desirable to discuss the relation between MTF and imagery quality. With regard to high spatial resolution remote sensing imagery, MTF mainly affects edge sharpness of imagery. And edge sharpness of imagery can affect imagery segmentation quality. So, the goal of this study is to analyze the effect of MTF on high spatial resolution remote sensing imagery segmentation quality. The paper firstly introduced the algorithm and steps of measuring MTF value based on high spatial resolution remote sensing imagery. Next, a typical imagery, which was derived from an acquisition over Florida, was selected for calculating MTF. In order to simulate different MTF values, the Gaussian PSF (Point Spread Function) was artificially added into the selected imagery. As a result, we can acquire a series of images, which had different MTF values. Then, the images were segmented by the object-based image analyst tool, such as eCongniton Developer 8.64, Feature Analyst, ENVI FX, et al. Finally, we compared the different segmentation quality by the methods of the segmentation accuracy assessment. The Area-Fit-Index (AFI) and Offspring-Loyalty (OL) were used to assess the segmentation accuracy of the images. The paper randomly selected 410 segmented polygons for assessing segmentation accuracy. The results of AFI and OL showed that the segmentation accuracy was lower while MTF value was lower. While MTF dropped still further, the AFI became negative number, −0.04. This showed that the edge profiles of the segmented objects mostly exceeded the edge profiles of the reference objects. The research also demonstrated that different land types had different segmentation quality on the condition of the same MTF value.

Optical remote sensing images were prone to extensive cloud coverage, especially under mountainous conditions with frequent weather changes. To address this issue, this paper proposed a novel cloud removal method that integrated features from SAR and neighboring optical remote sensing images. The method was based on a deep CGAN network, leveraging both local and global features of SAR images as well as edge features of optical remote sensing images to perform coarse cloud removal. Building upon the coarse cloud removal, the spectral features of neighboring optical remote sensing images were utilized for refined cloud removal. The experimental results showed that the proposed coarse-to-fine cloud removal method achieved a satisfactory cloud removal performance for optical remote sensing images. The RMSE, SAM, mSSI, and CC values were 0.0391, 0.0729, 0.9221, and 0.9537, respectively. Compared with other classical methods, each of these four metrics improved by at least 0.0119, 0.0438, 0.0217, and 0.0240, respectively. Moreover, the proposed method demonstrated superior performance in terms of boundary smoothness, cloud shadow elimination, spectral consistency, and spatial detail restoration for cloud removal in images with different underlying surface conditions. The effectiveness of the proposed method will improve the quality of cloud removal in optical remote sensing images.

As a key tectonic zone formed by the India–Eurasia collision, the eastern Tibetan Plateau has complex crustal structures and intense tectonic activity. Its crustal stability is closely linked to regional geohazards and the safety of major engineering projects. This study assessed crustal stability using nine multi-source datasets, including terrestrial gravity, Bouguer gravity anomalies, seismic data, active faults, terrain indicators, and annual precipitation. A hybrid weighting framework coupling the Analytic Hierarchy Process (AHP) and CRITIC method was established, and factor analysis was further adopted to cross-verify the rationality of index weights. Moderately unstable and unstable zones appear as alternating bands with distinct linear extensions. These unstable areas are mainly distributed around Jiuquan–Zhangye–Wuwei, Xining–Haidong, Yushu–Garzê, Mianyang, and Chengdu. Statistically, stable, moderately stable, moderately unstable, and unstable zones account for 20.74%, 35.47%, 28.78%, and 15.01%, respectively. The results provide a scientific reference for regional planning and infrastructure site selection in tectonically active regions.

Rice is one of the most important staple crops in China, and the rapid and accurate extraction of rice planting areas plays a crucial role in the agricultural management and food security assessment. However, the existing rice field identification methods faced the significant challenges in mountainous regions due to the severe cloud contamination, insufficient utilization of multi-dimensional features, and limited classification accuracy. This study presented a novel rice field identification method based on the Graph Convolutional Networks (GCN) that effectively integrated multi-source remote sensing data tailored for the complex mountainous terrain. A coarse-to-fine cloud removal strategy was developed by fusing the synthetic aperture radar (SAR) imagery with temporally adjacent optical remote sensing imagery, achieving high cloud removal accuracy, thereby providing reliable and clear optical data for the subsequent rice mapping. A comprehensive multi-feature library comprising spectral, texture, polarization, and terrain attributes was constructed and optimized via a stepwise selection process. Furthermore, the 19 key features were established to enhance the classification performance. The proposed method achieved an overall accuracy of 98.3% for the rice field identification in Huoshan County of the Dabie Mountains, and a 96.8% consistency compared to statistical yearbook data. The ablation experiments demonstrated that incorporating terrain features substantially improved the rice field identification accuracy under the complex topographic conditions. The comparative evaluations against support vector machine (SVM), random forest (RF), and U-Net models confirmed the superiority of the proposed method in terms of accuracy, local performance, terrain adaptability, training sample requirement, and computational cost, and demonstrated its effectiveness and applicability for the high-precision rice field distribution mapping in mountainous environments.