A Landslide Detection Method Based on CBAM Attention Mechanism Optimized for YOLOv8n

Accurate landslide detection is crucial for disaster prevention and early warning.Existing detection methods often suffer from interference by complex backgrounds and small-target features,particularly in high-resolution remote sensing imagery.To address these challenges,this study proposes an enhanced YOLOv8n model incorporating the Convolutional Block Attention Module (CBAM) for landslide detection in high-resolution remote sensing images (YOLOv8n-CBAM),validated using high-resolution remote sensing datasets.The model demonstrates three key advantages:(1)It significantly enhances attention to critical features in complex terrains and vegetation-covered scenarios,improving detection accuracy for small landslides while reducing missed and false detections.(2)Evaluation metrics including normalized confusion matrix,precision-recall curves,and loss curves confirm superior detection capability,accuracy,and robustness compared to the baseline YOLOv8n model.(3)Comparative analysis reveals optimal performance in precision (89.7%),recall (86.4%),mAP@0.5 (90.1%),and mAP@0.5:0.95 (67.3%) metrics among state-of-the-art detection methods.