Extracting farmland from high spatial resolution remote sensing images is a basic task for agricultural information management. According to Tobler’s first law of geography, closer objects have a stronger relation. Meanwhile, due to the scale effect, there are differences on both spatial and attribute scales among different kinds of objects. Thus, it is not appropriate to segment images with unique or fixed parameters for different kinds of objects. In view of this, this paper presents a stratified object-based farmland extraction method, which includes two key processes: one is image region division on a rough scale and the other is scale parameter pre-estimation within local regions. Firstly, the image in RGB color space is converted into HSV color space, and then the texture features of the hue layer are calculated using the grey level co-occurrence matrix method. Thus, the whole image can be divided into different regions based on the texture features, such as the mean and homogeneity. Secondly, within local regions, the optimal spatial scale segmentation parameter was pre-estimated by average local variance and its first-order and second-order rate of change. The optimal attribute scale segmentation parameter can be estimated based on the histogram of local variance. Through stratified regionalization and local segmentation parameters estimation, fine farmland segmentation can be achieved. GF-2 and Quickbird images were used in this paper, and mean-shift and multi-resolution segmentation algorithms were applied as examples to verify the validity of the proposed method. The experimental results have shown that the stratified processing method can release under-segmentation and over-segmentation phenomena to a certain extent, which ultimately benefits the accurate farmland information extraction.

Automatic landslide detection based on very high spatial resolution (VHR) remote sensing images is crucial for disaster prevention and mitigation applications. With the rapid development of deep learning techniques, state-of-the-art semantic segmentation methods based on fully convolutional network (FCNN) have achieved outstanding performance in the landslide detection task. However, most of the existing studies only utilize visual features. Even if the advanced FCNN models are applied, there is still a certain amount of falsely detected and miss detected landslides. In this article, we innovatively introduce landslide susceptibility as prior knowledge and propose an innovative susceptibility-guided landslide detection method based on FCNN (SG-FCNN) to detect landslides from single temporal images. In addition, an unsupervised change detection method based on the mean changing magnitude of objects (MCMO) is further proposed and integrated with the SG-FCNN to detect newly occurred landslides from bi-temporal images. The effectiveness of the proposed SG-FCNN and MCMO has been tested in Lantau Island, Hong Kong. The experimental results show that the SG-FCNN can significantly reduce the amount of falsely detected and miss detected landslides compared with the FCNN. It can conclude that applying landslide susceptibility as prior knowledge is much more effective than using visual features only, which introduces a new methodology of landslide detection and lifts the detection performance to a new level.

The spatial arrangement of land-cover features constitutes different urban functional zone. With the same attributes of the urban functional zone, the land-cover features that make up the functional zone will have similar spatial distribution characteristics. Considering the importance of understanding spatial relationships between land-cover features, the up-bottom hierarchical decomposition and semantic understanding of functional zone are achieved. First, for object convolution neural network (OCNN)-based land cover classification, an equal-area dividing algorithm is proposed to automatically generate convolution kernel position. Second, to extract spatial relationship features of urban land covers, a novel distance weight-graph attention model (DW-GAM) is originally proposed for classifying urban functional zones by comparing the feature similarity of the land cover relationship graph. Third, considering the extreme difficulties in expressing the urban structure characteristic on a single scale, a recursive model that uses an urban road network of different levels to divide multiscale functional zones is built. Finally, taking the analysis of urban function allocation as the application objective, this article establishes a primary framework of the spatial pattern evaluation index. Experimental results conducted on a Google Earth image of Xi’an city show that the multiscale recursive model can accurately recognize urban functional zones by using the originally proposed DW-GAM. Then, based on the outcome of urban functional zone identification, the case study of urban function allocation analysis is innovatively conducted on the fine scale to give some suggestions on future urban planning, which is, hence, of great significance for urban function pattern analysis and urban planning.

The advent of remote sensing and deep learning has rendered the automated detection of post-earthquake building damage a necessity for effective disaster response. However, the inherent unpredictability of earthquakes and the limited revisit frequency of satellites frequently result in inadequate post-disaster data for effective model training. Furthermore, conventional data augmentation techniques have the potential to lead to model overfitting, thereby further complicating the development of robust detection systems.To address the challenge, this paper introduces an innovative end-to-end data augmentation approach that combines a Multi-Channel Fusion Wasserstein Generative Adversarial Network (MCFWGAN) with a Mask-CutMix sample augmentation technique. This approach first introduces a Multi-Channel Fusion (MCF) mechanism into the WGAN framework, constructing the MCFWGAN model, which enhances global feature extraction capabilities while strengthening the interdependencies among image channels. After generating images, the Mask-CutMix method integrates these generated images into large-scale remote sensing imagery, thereby augmenting the dataset.The integrated image data set contains a small amount of pre- and post-earthquake data.Finally, a target detection network is employed to extract damaged buildings from the images. Two study areas were selected to implement data augmentation and target detection. Experimental results demonstrate that the proposed data augmentation method outperforms traditional techniques in terms of accuracy metrics for target detection. Furthermore, to address the issues of missed and false detections of intact buildings within severely damaged areas, the use of style transfer-based sample augmentation was shown to effectively improve accuracy metrics.The experimental results demonstrate that the proposed data enhancement method outperforms the traditional technique with respect to the target detection evaluation metric. In Study Area 1, the detection accuracy is enhanced by 5.08% in comparison to the conventional data enhancement method. In Study Area 2, the mean accuracy is elevated by 3.32%. The data augmentation method proposed in this study not only increases the number of samples but also enhances the diversity of samples representing damaged buildings. The MCFWGAN model is particularly well-suited for generating images of damaged buildings in complex post-earthquake scenarios, offering significant potential for the accurate acquisition of post-disaster information and emergency rescue operations.

Currently in Landslide Susceptibility Mapping (LSM), data-driven models mostly utilize association to present the connection between landslide and Conditioning Factor (CF). However, association cannot reflect the structural interaction or the underlying causal mechanism. This leads to black-box results and reduces interpretability. To address this problem, this study introduces an unsupervised causal model to LSM. The methodology consists of two parts: causal discovery with Non-combinatoric Optimization via Trace Exponential Augmented Lagrangian Structure learning (NOTEARS) algorithm, and causal inference with Bayesian Networks (BN). By applying structural causality to LSM, we “white-boxed” the procession of data-driven model, thereby ensuring the interpretability and data utility. To comprehensively evaluate the results, comparison experiments with other machine learning models were performed. The results showed that BN achieved the highest accuracy (approximately 0.877) and the second highest AUC (0.907). Moreover, BN outperformed other models in ground-truth validation. The high and extremely high susceptibility zones in BN-LSM result covered the most actual landslides (more than 71.07% of the total, and more than 80% after excluding areas not used in model training). This showed the generalizability and stability of structural causality-guided-BN. The proposed methodology provides a link between data-driven causal science and natural hazard analysis, which can help intelligent hazard managements. Data and codes that support the findings of this study are openly available at doi.org/10.6084/m9.figshare.26395420.