关键词: YOLOv8_n, 路面缺陷检测, 长程依赖关系, 裂缝, 多尺度

Abstract: To address the problem of low detection accuracy and high missed detection rates caused by the narrow lateral, multi-scale, and long-range dependency characteristics of Pavement defects, a road defect detection algorithm based on the improved YOLOv8_n with enhanced morphological perception was proposed. First, an Edge-Enhancement Focus Module (EEFM) was introduced in the backbone network fusion stage. A strip pooling kernel was used to capture directional and position-aware information, enhancing edge details in deep features and improving the expression of elongated features. Second, a Dual Chain Feature Redistribution Pyramid Network (DCFRPN) was designed to reconstruct the fusion method, providing multi-scale features with extensive perceptual and rich localization information, improving the fusion ability for multi-scale defects. Additionally, a Morphological Aware Task Interaction Detection Head (MATIDH) was constructed to enhance task interaction between classification and localization, dynamically adjusting data representation and integrating multi-scale strip convolutions to optimize the classification and regression of elongated defects. Finally, a PWIoU (Penalized Weighted Intersection over Union) loss function was proposed to dynamically allocate gradient gains for prediction boxes of different qualities, optimizing the regression of bounding boxes. Experimental results showed that, on the RDD2022 dataset, compared to the baseline, precision and recall were improved by 3.5 and 2.3 percentage points, respectively, and mean average precision at 50% IoU increased by 3.2 percentage points, verifying the effectiveness of the algorithm. 

Key words: YOLOv8_n, pavement defect detection, long-range dependency, crack, multi-scale