The region merging algorithm is a widely used segmentation technique for very high resolution (VHR) remote sensing images. However, the segmentation of post-earthquake VHR images is more difficult due to the complexity of these images, especially high intra-class and low inter-class variability among damage objects. Herein two key issues must be resolved: the first is to find an appropriate descriptor to measure the similarity of two adjacent regions since they exhibit high complexity among the diverse damage objects, such as landslides, debris flow, and collapsed buildings. The other is how to solve over-segmentation and under-segmentation problems, which are commonly encountered with conventional merging strategies due to their strong dependence on local information. To tackle these two issues, an adaptive dynamic region merging approach (ADRM) is introduced, which combines an adaptive spectral-spatial descriptor and a dynamic merging strategy to adapt to the changes of merging regions for successfully detecting objects scattered globally in a post-earthquake image. In the new descriptor, the spectral similarity and spatial similarity of any two adjacent regions are automatically combined to measure their similarity. Accordingly, the new descriptor offers adaptive semantic descriptions for geo-objects and thus is capable of characterizing different damage objects. Besides, in the dynamic region merging strategy, the adaptive spectral-spatial descriptor is embedded in the defined testing order and combined with graph models to construct a dynamic merging strategy. The new strategy can find the global optimal merging order and ensures that the most similar regions are merged at first. With combination of the two strategies, ADRM can identify spatially scattered objects and alleviates the phenomenon of over-segmentation and under-segmentation. The performance of ADRM has been evaluated by comparing with four state-of-the-art segmentation methods, including the fractal net evolution approach (FNEA, as implemented in the eCognition software, Trimble Inc., Westminster, CO, USA), the J-value segmentation (JSEG) method, the graph-based segmentation (GSEG) method, and the statistical region merging (SRM) approach. The experiments were conducted on six VHR subarea images captured by RGB sensors mounted on aerial platforms, which were acquired after the 2008 Wenchuan Ms 8.0 earthquake. Quantitative and qualitative assessments demonstrated that the proposed method offers high feasibility and improved accuracy in the segmentation of post-earthquake VHR aerial images.

In recent years, the use of street-view images for urban analysis has received much attention. Despite the abundance of raw data, existing supervised learning methods heavily rely on large-scale and high-quality labels. Faced with the challenge of label scarcity in urban scene classification tasks, an innovative self-supervised learning framework, Trilateral Redundancy Reduction (Tri-ReD) is proposed. In this framework, a more restrictive loss, “trilateral loss”, is proposed. By compelling the embedding of positive samples to be highly correlated, it guides the pre-trained model to learn more essential representations without semantic labels. Furthermore, a novel data augmentation strategy, tri-branch mutually exclusive augmentation (Tri-MExA), is proposed. Its aim is to reduce the uncertainties introduced by traditional random augmentation methods. As a model pre-training method, Tri-ReD framework is architecture-agnostic, performing effectively on both CNNs and ViTs, which makes it adaptable for a wide variety of downstream tasks. In this paper, 116,491 unlabeled street-view images were used to pre-train models by Tri-ReD to obtain the general representation of urban scenes at the ground level. These pre-trained models were then fine-tuned using supervised data with semantic labels (17,600 images from BIC_GSV and 12,871 from BEAUTY) for the final classification task. Experimental results demonstrate that the proposed self-supervised pre-training method outperformed the direct supervised learning approaches for urban functional zone identification by 19% on average. It also surpassed the performance of models pre-trained on ImageNet by around 11%, achieving state-of-the-art (SOTA) results in self-supervised pre-training.

Transformer-based methods have achieved remarkable progress in single-image deraining through long-range dependency modeling. However, they remain limited by redundant responses from dense channel interactions, insufficient characterization of rain-induced frequency degradation, and limited adaptability of U-shaped decoders to feature fusion and reconstruction. To address these issues, we propose ASFANet, an Asymmetric Spatial-Frequency Sparse Attention Network for image deraining. First, we develop Dynamic Sparse Channel Attention (DSCA) for adaptive feature selection. Its Spatial-Aware Predictor (SAP) adaptively estimates the proportion of channel responses to retain, while a Straight-Through Estimator (STE) provides surrogate gradients for hard top-k selection, enabling end-to-end optimization. Second, we introduce a Spatial-Frequency Feed-forward Network (SFFN) that combines spatially gated modeling and frequency modulation to capture local textures and rain-related frequency responses. Finally, we construct an asymmetric decoder with Upsampling Gated Fusion (UGF) and an Adaptive Dual-domain Gated Block (ADGB). It adaptively modulates and fuses skip features and performs spatial-frequency guided reconstruction to improve feature utilization and detail restoration. Experiments on four public benchmarks, including Rain200H, demonstrate superior deraining performance, with PSNR gains of 0.10–0.76 dB over Restormer. Further ablations show that the asymmetric decoder maintains restoration performance while reducing parameters by 4.55 M compared with the original homogeneous Transformer decoder.