Aiming at the problem that it is difficult to directly detect bridge apparent cracks under com-plex background conditions,this paper proposes a bridge apparent crack detection algorithm based on deep learning.Firstly,the collected bridge apparent crack image is divided into small-sized bridge crack patches and bridge background patches by sliding window algorithm,and a bridge crack classifi-cation model based on Inception network and residual network(ResNet)is proposed according to the analysis of the patches,which is used to identify bridge crack patches and bridge background patches.Then,the bridge crack classification model and sliding window algorithm are combined to detect the apparent crack image of the bridge.Finally,the width of the crack is measured using digital image processing technology.The experimental results show that the algorithm in this paper has more than 99％classification accuracy for the apparent cracks of the bridge,which can meet the actual engineer-ing needs.The extraction of cracks is realized and the position of cracks in the image can be accurately located.The crack width is measured according to the imaging principle.Compared with the tradition-al deep learning model,the model has higher execution efficiency,can be used for large-scale detec-tion,and is easier to apply in bridge health detection.
NETL
NSTL
万方数据
