Pavement Crack Recognition Based on ResNet and Attention Mechanism
In order to solve the problems of insufficient feature attention and interference and noise problems such as lighting,scale and occlusion in pavement crack image recognition,an improved ResNet pavement crack classification method is proposed.Firstly,the basic ResNet18 is used as the backbone network to extract the pavement crack features,and secondly,the attention module is in-troduced after each residual block of the basic model to guide the network to pay attention to the crack area.The self-built pavement crack dataset was used to compare the proposed model with mul-tiple mainstream models,and the models before and after optimization were compared with each oth-er.Experimental results show that the accuracy,recall and precision of the proposed model on the self-built pavement crack dataset are higher than those of the baseline model,reaching 94.30％,94.33％and 94.49％,respectively,and the crack classification results are accurate and effective,and the model can meet the needs of pavement crack classification without human intervention.
万方数据
