The Research On Key Technology Of Car Surface Scratch Detection System

The number of imported cars in China has been increasing rapidly,and a large number of imported cars need to be detected for surface scratches before entering warehouse.At present,the surface scratches detection of car is detected by manual detection,which leads to low efficiency,high work intensity and high undetected rate.Therefore,computers are urgently needed to replace manual detection.The surface scratch detection system based on image processing is designed in this thesis.In order to meet high demand of accuracy,efficiency and reliability,the key techniques of edge detection and surface scratch location of cars are studied in this thesis.The main research contents are as follows:(1)The requirement of car surface scratch detection system is analyzed.The hardware and the design of the overall framework and the overall process are given,and key technique of the system is introduced in this thesis.(2)The technique of car edge detection is studied.Aiming at the edge characteristics of cars,the SUSAN operator is extended to the edge detection of CIELab color space and directional SUSAN operator,and a car edge detection algorithm based on improved SUSAN operator is designed.The algor ithm improves the decision function of the SUSAN operator's USAN,and extends the application of the chromaticity information.Then,the scanning window of SUSAN algor ithm was improved by adding directivity feature to improve the accuracy of edge detection and the purpose of edge continuity.The experiment shows that the technology meets the requirement of the edge detection accuracy and high efficiency.(3)The technique of template matching,image registration and surface scratch location is studied.Aiming at the characteristics of car surface scratches,a template matching and image registration algor ithm is proposed.The algorithm improved the matching error calculation method of SSDA,reduced the mismatching caused by the linear change of grayscale,and improved the accuracy of SSDA template matching.Then the image registration space transformation parameter calculation method is improved to improve the speed of image registration and scratch location.Experiments show that the technology can accurately and accurately locate the car surface.(4)The car surface scratch detection system is realized.The selection of the main devise models of the hardware is completed.The main functions of the system are realized and the main functions operation examples of the system are given in this thesis.