Study Of Surface Defects Automatic Detection System Based On Image Processing Technology

Brake valve, one of the components of braking system that plays an importantrole on security of vehicles, is closely concerned with reliability, stability and securityof braking system. The brake valves with defects on the working face must be foundand eliminated before they are installed into automobiles to keep from suddenaccidents. Traditional detection methods adopting sample and manual eyeballing giverise to immense labour intensity and high ratio of wrong detection, which limitimprovement of productivity and quality. The dissertation introduces an automaticdetection system of defect on the working face.The work can be included as follows:1) The hardware system for imaging is designed on the base of analyzing andsumming up the three characteristics of working face.2) Some kinds of image processing and recognition technologies are utilized toacquire characteristic parameters of defects, such as histogram equilibrium,improved filter technology combining mean filter and median filter, automaticthreshold selection using moment invariant features, extraction of working faceusing Hough transform, mathematical morphology, regional labeling.3) Defect classes and grades are defined according to shape of defects andinfluence on braking performance, respectively. The output of automaticdetection's result not only shows working face defect's classes and grades, butalso details of characteristic parameters of defects are saved as a document inthe purpose of authentication and statistics in the future.4) The software system of automatic detection of working face defect isimplemented in visual c++ environment.The major innovations achieved in the dissertation are as follows:1) Improved filter technology combining mean filter and median filter is utilizedwith the filter coefficient adjusted to the character of the acquired image.2) Working face is extracted from the whole image with Hough transform and thecomputational complexity of Hough transform is decreased in the light of thedimensional orientation character of the acquired image.3) Defect classes and grades are defined according to shape of defect andinfluence on braking performance, respectively. The quantitative standard ofDefect classes and grades is presented tentatively.Running in the product line shows that this system with characteristic ofnon-contact and automatism can work steadily and obtain accurate detection results,which can never be matched by the manual detection.