A cross‐environmental algorithm for multi‐scale road crack detection under extreme weather conditions

This study addresses road crack detection challenges in extreme weather conditions by introducing the cascaded cross‐scale multibranch detection transformer (CCM‐DETR) framework. The research advances three key scientific contributions: (1) a specialized extreme weather road crack dataset; (2) a novel detection architecture CCM‐DETR with triple optimization strategy for efficient feature processing; and (3) a cross‐environment stability evaluation method. Validated on extreme weather conditions (varying fog densities, rainfall intensities, and snow coverage), the model achieves 51.4% average precision across intersection over union (IoU) thresholds of 0.5–0.95 (AP 50‐95 ) and 89.7% AP at IoU threshold 0.5 (AP 50 ), with balanced performance across object scales (AP s : 38.5%, AP m : 51.7%, and AP l : 62.0%). Containing only 17.4 M parameters and 65.3 G floating point operations, CCM‐DETR demonstrates remarkable robustness—experiencing merely 0.2% performance fluctuation in extreme weather versus 8.5% degradation in baseline models. This enables reliable crack detection where traditional methods fail, advancing intelligent transportation infrastructure inspection.