Research on automatic pavement crack identification Based on improved YOLOv8

Pavement crack detection is an important task in the periodic inspection of pavements. Aiming at the problems of high labor cost and low recognition rate during traditional pavement crack detection, this paper proposes an improved automatic recognition algorithm based on YOLOv8 model, which is defined as YOLOv8-D-CBAM. First, in order to flexibly capture the crack informationand improve the model recognition rate. Depthwise separable convolution is introduced to the backbone network of the model; Second, Convolutional Block Attention Module (CBAM) is added to the neck to improve the model feature extraction capability. Improved model recognition rate and model performance are achieved based on the above improvements. Experimental results show that the YOLOv8-D-CBAM algorithm of Precision, Recall, F1-score and mAP@0.5 are 98.21%, 98.21%, 97.9% and 99.5% respectively; Meanwhile, The model achieved more than 90% recognition rate for each category of crack disease. So the YOLOv8-D-CBAM algorithm has good model performance in pavement crack identification research.