An Integrated Method for Road Crack Segmentation and Surface Feature Quantification under Complex Backgrounds

In the present study, an integrated framework for automatic detection, segmentation, and measurement of road surface cracks is proposed. First, images are captured, crack regions detected based on fifth version You Only Look Once (YOLOv5) algorithm; then, a modified Residual Unity Networking (Res-UNet) algorithm proposed accurate segmentation at pixel level within regions; finally, novel feature quantification developed to determine pixels in width length, respectively. addition, dataset containing complex environmental noise produced. Different shooting distances, angles, lighting conditions considered. Validated through same compared with CoefficienTs ++ (YOLACT++) DeepLabv3+, method shows higher accuracy under backgrounds. Specifically, damage detection YOLOv5 achieves mean average precision 91%; Res-UNet 87% intersection over union (IoU) when segmenting pixels, 6.7% than original Res-UNet; has 95% identifying length root square error 2.1 width, being 3% that traditional method.