Road crack detection algorithm based on YOLOv3

In order to overcome the problems that may cause pedestrians, driving safety and other major economic losses, road cracks must be discovered in time and resolved as soon as possible. The paper proposes the detection of road cracks based on the YOLOv3 algorithm. In order to further improve the accuracy, this method is optimized on the basis of the traditional YOLOv3 algorithm, and uses MSE, GIOU_Loss and CIOU_Loss as the bounding box regression loss function into YOLOv3 to detect road cracks. Experimental results show that the improved detection accuracy of YOLOv3 algorithm using MSE, GIOU_Loss and CIOU_Loss loss functions are 84.98%, 88.73% and 92.97%, respectively. It can be seen that the YOLOv3 algorithm using the CIOU_Loss loss function can identify cracks more quickly and accurately while maintaining real-time performance. This method provides a simpler, more intuitive and effective way for road health. It has certain value in real life applications.