Research And Implementation Of Pavement Crack Detection Method Based On Deep Learning

In road maintenance,the detection of pavement damage is the primary link of road maintenance,and pavement crack is the main form of pavement damage.The detection of pavement crack can enable the road administration to keep abreast of road damage and provide decision-making basis for formulating road maintenance plans and measures.In this paper,pavement crack detection is chosed as the object to conduct related research on pavement image preprocessing,pavement crack detection algorithm and classification for degree of pavement damage,and the target detection algorithm based on deep learning is adopted.Firstly,according to the characteristics of the pavement crack image,Gaussian bilateral filtering is used to remove the noise of pavement image while retaining the information of crack edge.In order to solve the problem of insufficient number of collected road crack pictures,the data of crack image samples were expanded by means of image flipping,translation and color adjustment.Then,the paper compares the R-CNN series algorithm and YOLO series algorithm of deep learning in the field of target detection,expounds the advantages of Mask R-CNN algorithm for pavement crack detection,and optimizes the training and testing of Mask R-CNN algorithm.The experimental results show that the improved Mask R-CNN algorithm achieves accuracy of 0.9412,recall rate of 0.9143,F1 value of 0.9275 and mAP value of 0.9082 in the test set,and the detection effect is good.Finally,based on the detection of pavement crack by MASK R-CNN,the paper further extracts 14 pavement crack characteristics and constructs a random forest model to classify the degree of pavement damage.The experimental results show that the accuracy of the random forest model reaches 0.9221,the recall rate reaches 0.8943,and the F1 value reaches 0.9080 in the test set,and the classification effect is good.In summary,this paper studies and implements pavement crack detection based on deep learning algorithm,which provides a new method for pavement crack detection and data support for further maintenance decision-making.