Research On Pavement Defect Detection Method Based On Deep Learning

At present,the volume of road facilities in my country is increasing year by year.In order to truly quantify the degree of road damage,improve the level of road maintenance,and ensure safe,reliable and fast transportation processes,it is necessary to select reasonable and efficient detection methods.However,in the current mainstream maintenance process,it is usually done manually,which not only requires a lot of time and energy,but also has low work efficiency.Therefore,realizing accurate identification and automatic detection of pavement defects is a practical engineering problem that needs to be solved urgently.The thesis takes the road surface defect detection as the research content,and focuses on the road image preprocessing and the road surface defect detection algorithm.It designs a method to automatically detect the defects in the road surface image based on the Mask RCNN algorithm,and develops the corresponding visualization interface.First,according to the characteristics of the road surface defect image,the paper uses histogram equalization on the road surface image to enhance the image contrast,and uses bilateral filtering to remove the road surface image noise while retaining the edge information of the defect.The pavement defect image is enhanced and the sample is expanded to solve the problem of insufficient number of pavement defect images.Second,the paper describes the basic knowledge of deep learning and the overall framework of convolutional neural networks,analyzes the structure and principles of classic deep learning algorithms,expounds the role of Res Net,analyzes the Mask RCNN algorithm,and optimizes the training and testing of the Mask RCNN algorithm.Experimental results show that the trained model can automatically learn the effective features of each road surface defect image and realize automatic classification of road surface defect images.The average accuracy of the average value reaches 90.42%,and the detection effect is good.Finally,the paper obtained a better-performing pavement defect detection model,developed a human-computer interaction interface,and built a pavement defect detection platform based on deep learning.Through the research of this paper,a new method is provided for the detection of pavement defects,which realizes the automatic detection of pavement defects,and further promotes the development of pavement engineering maintenance detection technology.