PCLCD: Patch-Based Contrastive Learning for Remote Sensing Image Change Detection

Current remote sensing change detection remains a challenging task due to severe pseudo-change interference and insufficient local feature modeling in complex scenarios. To address these issues, we propose a novel Patch-based Contrastive Learning Change Detection framework (PCLCD), which systematically enhances discriminative feature representation through bi-temporal image pair analysis. The core innovation involves two modules: 1) A Patch-based Contrastive Learning Module (PCLM) that explicitly captures inter-patch similarity relationships by generating dense feature-domain patches. This module strengthens feature consistency within homogeneous regions via positive sample pairs while suppressing semantic confusion in non-change areas through negative pair contrastion; and 2) A Dynamic Contextual Aggregation Module (DCAM) that adaptively fuses multi-scale local patterns with global contextual information, enabling robust change discrimination across varying spatial scales. Extensive experiments on benchmark datasets demonstrate that PCLCD achieves state-of-the-art performance, with significant improvements on LEVIR-CD, WHU-CD, and SYSU-CD.