Deep Learning-based Crack Detection Of Highway Pavement

As the investment in highway construction continues in china has increased year by year,road maintenance has become more and more important.In order to the safety of vehicles on the road,The detection of pavement crack can enable the road managerment department keeps track of the damage of the road in time and enable take measures to repair roads in time.In recent years,deep learning has been rapidly developed in the field of computer vision,and the accuracy of deep learning recognition has improved a lot.The greatest strength of deep learning does not require manual design features,and can automatically abstract expression based on the original image features.According to the problem of crack detection on highway pavement,a method for detecting cracks in road images by convolutional neural network is designed for pavement crack detection.The main contents of this paper include:First,introduced the basic theories and principles of deep learning convolutional neural networks,the composition of convolutional neural networks and the role of each layer.The paper introduces common target detection algorithms,Contrast the R-CNN algorithm with YOLO algorithm.Recently,the Yolo5 has good performance in target detection speed and accuracy.Finally,introduce the network structure of Yolo5.Second,in order to remove the noise of pavement image.In term of the feature of the pavement crack image,Using Gaussian bilateral filtering to obtain the key feature information of image.Deep learning need a lot of data.The number of collected road crack pictures are not enough,so increase the number of road cracked picture by flipping and panning the picture.And label the images of the dataset.Finally,the paper detection of pavement crack by YOLO5,optimizes the training and testing of YOLO5 algorithm,in the the road crack pictures with complex background,The algorithm has achieved good results for the road crack target recognition location and the crack classification.In summary,this paper detection of pavement crack by optimize the YOLO5,which provides a fast and reliable road crack detection method,which has an engineering application value.