Design Of Pavement Crack Recognition System Based On SVM And Convolutional Neural Network

The road crack is the defect which mainly causes road safety problems,and the detection and maintenance of road cracks are effective measures to solve these problems.raditional manual detection costs high in manpower and material resources,with low efficiency in comprehensive detection.Therefore,the research of road crack detection method and the design of monitoring system have important practical significance and high practical value.On the basis of consulting and analyzing the relevant domestic and foreign references of the topic,the main hardware such as road crack image recognition algorithm,background processing software platform and image acquisition unit are compared and selected in the thesis,to determine the design of overall scheme of road crack recognition system.Then,the image recognition algorithm based on support vector machine(SVM)and convolutional neural networks(CNN)is analyzed and optimized for classification recognition accuracy.Second,the hardware equipment is selected according to the function and performance requirements of the road crack recognition system,and the acquisition test platform for road crack recognition is constructed by installing camera,image acquisition card and auxiliary light source at the rear of the car.Third,Halcon image processing software and Keras machine learning framework are used for designing the road crack recognition and background image processing software.Finally,the system test and test results are compared and analyzed by the designed software.The test results show that the accuracy of optimized algorithm is more than 94%,meeting the requirements of engineering practice.The identification system designed in this thesis costs less,with high efficiency,convenient installation and wide application in engineering practice.