A Cultivated Land Change Detection Method Based on Dual-branch Spatial Channel Attention

Cropland is a fundamental resource for agricultural production, and monitoring its changes is crucial for effective management. This paper introduces the Dual-branch Spatial Channel Attention Network (DBSCANet) for cropland change detection with high-resolution remote sensing images. DBSCANet leverages a pre-trained HRNet to extract multiscale cropland features, ensuring robust representation. A dual-branch spatial channel attention (DBSCA) module is designed for undirected change feature extraction, complemented by a hybrid symmetry loss function that enhances the learning of precise change information. Experiments conducted on the CLCD dataset demonstrate the superior performance of DBSCANet. Ablation studies confirm the effectiveness of the DBSCA module, and comparative analyses highlight DBSCANet's competitive balance of high accuracy and computational efficiency, making it well-suited for practical applications.