A Spatial Semantics and Continuity Perception Attention for Remote Sensing Water Body Change Detection

TL;DR

This work tackles Water Body Change Detection (WBCD) in high-detail remote sensing by releasing the HSRW-CD dataset (spatial resolution better than 3 meters) and introducing the Spatial Semantics and Continuity Perception (SSCP) attention module. SSCP consists of three components—Multi-Semantic Spatial Attention (MSA), Structural Relation-aware Global Attention (SRGA), and Channel-wise Self-Attention (CSA)—and can be plugged into existing WBCD models to strengthen spatial semantics and structural continuity in deep features. Comprehensive experiments on HSRW-CD and Water-CD show that SSCP yields consistent gains in F1 and IoU, with ablations confirming the contribution of each sub-module and demonstrating robustness across datasets. The approach advances practical WBCD for urban/rural water-resource management by enabling more accurate delineation of water body changes in complex scenes. Future work includes developing a WBCD-specific network that further leverages multi-scale features to mitigate challenging failure cases.

Abstract

Remote sensing Water Body Change Detection (WBCD) aims to detect water body surface changes from bi-temporal images of the same geographic area. Recently, the scarcity of high spatial resolution datasets for WBCD restricts its application in urban and rural regions, which require more accurate positioning. Meanwhile, previous deep learning-based methods fail to comprehensively exploit the spatial semantic and structural information in deep features in the change detection networks. To resolve these concerns, we first propose a new dataset, HSRW-CD, with a spatial resolution higher than 3 meters for WBCD. Specifically, it contains a large number of image pairs, widely covering various water body types. Besides, a Spatial Semantics and Continuity Perception (SSCP) attention module is designed to fully leverage both the spatial semantics and structure of deep features in the WBCD networks, significantly improving the discrimination capability for water body. The proposed SSCP has three components: the Multi-Semantic spatial Attention (MSA), the Structural Relation-aware Global Attention (SRGA), and the Channel-wise Self-Attention (CSA). The MSA enhances the spatial semantics of water body features and provides precise spatial semantic priors for the CSA. Then, the SRGA further extracts spatial structure to learn the spatial continuity of the water body. Finally, the CSA utilizes the spatial semantic and structural priors from the MSA and SRGA to compute the similarity across channels. Specifically designed as a plug-and-play module for water body deep features, the proposed SSCP allows integration into existing WBCD models. Numerous experiments conducted on the proposed HSRW-CD and Water-CD datasets validate the effectiveness and generalization of the SSCP. The code of this work and the HSRW-CD dataset will be accessed at https://github.com/QingMa1/SSCP.

Figures (11)

• Figure 1: The comparison of typical samples in Water-CD and HSRW-CD datasets. The top two rows present examples from the Water-CD dataset, while the following two rows of images come from HSRW-CD. It is clear that the proposed HSRW-CD has a higher spatial resolution than Water-CD. In addition, different water body types vary in shape and have strong spatial continuity.
• Figure 2: The overall structure of the proposed SSCP. Specifically, the deep feature map X from the CD networks passes through the MSA and SRGA module to extract spatial semantics and structural information. Finally, the CSA further deepens understanding by leveraging the spatial semantics and structural priors.
• Figure 3: The structure of change detection networks with SSCP. The SSCP module is utilized to process deep water body features.
• Figure 4: Samples of various water body types from the proposed HSRW-CD dataset. Each group of three consecutive columns represents sample images of a specific water body type from left to right.
• Figure 5: The test set visualization results of the HSRW-CD dataset compared with other advanced approaches. The red regions denote the false positive and the green regions are the false negative.