A Dual-Modal Framework Integrating SAR-Based Change Screening and Optical-Scene-Informed Identification for High-Frequency Monitoring of Construction-Ready Bare Land

Rapid urbanization necessitates high-frequency monitoring of construction-ready bare land to timely detect and prevent illegal construction. However, the utility of optical imagery is often compromised in cloud-prone regions. While Synthetic Aperture Radar (SAR) offers all-weather capabilities, it struggles to distinguish construction-ready bare land from recently harvested agricultural land, leading to severe false alarms. To address the conflict between high-frequency monitoring and semantic identification, this study proposes the SAR-based Change Screening and Optical-Scene-Informed Identification (SCS-OI) framework. The first stage performs high-recall candidate screening based on SAR backscattering changes, while the second stage incorporates historical cloud-free optical imagery as semantic guidance, enabling refined identification without requiring synchronous optical data. Experiments in Guangzhou demonstrate that the framework achieves a False Alarm Rate of 13.31%, Recall of 90.63%, Precision of 74.81%, F1-score of 81.95%, and IoU of 69.43%. Compared with the SAR-only baseline (FR = 22.4%), the two-stage design reduces false alarms while maintaining high recall. Other deep learning baselines exhibit lower F1-scores (59–73%), highlighting the effectiveness of the overall framework. These results show that the proposed two-stage framework effectively integrates high-recall candidate screening and semantic-guided refinement, providing a robust solution for high-frequency monitoring of construction-ready bare land in cloud-prone regions of Guangzhou.