Research On The Detection Of The Surface-deviation Of Door-outer Panel

Under the background of the rapid development of the domestic automobile industryatmosphere，the competition is more and more bigger in the car makers, Outside of the carbody covering parts requirements are put forward more strict around the car makers,thequality of the body outer covering parts is not only affecting the performance of the vehicle,but also having a great influence on the appearance of the car-body. So, we should assureboth the functional of the outer covering parts and the molding conditions, avoid surfacequality problems of the stamping parts, such as concave-convex, indentation, and wave. So,it is significant to detect the quality of the surface deviation on the stamped parts.The traditional deviation quality detecting on the surface of the outer cover parts isadopted by manual detecting, Rely on a certain experience of inspector for detecting bywearing cloth gloves to touch with and light reflection observation method, or adopting themethod of scratching detection. These detective method are fall within the scope ofqualitative,not be able to realize quantitative concept, so it is hard to be obtained for theseverity of the surface errors for the objective evaluation, the accuracy of the evaluation ofquality mainly depends on the quality inspection operation personnel’s professionaltechnology level, that is full of strong subjectivity. In these years, Binocular stereo visiontechnology can be used in the industrial quality detection as high speed and accuracy methodwith the rapid development of computer vision technology. And it can realize non-contacton-line quality detection, so the Binocular stereo vision technology is widely used inindustrial production.In this paper, we utilize the method of combining the Binocular stereo visiontechnology method to detect the surface deflection of the car–outer panel quantitatively,using the car-outer panel for routine experiment, Using a beam of laser light emitted on thecar–outer panel, the binocular stereo vision system is utilized to take the Photo of the laser,the picture processed in edge detection，stereo matching to restructure the3D information of the edge of the laser, so the ideal curve can be found by the least square circle method, thedeviation can be detected quantitatively by comparing with the ideal curve and the actualcurve. The main research content is as follows:1）The MATLAB camera calibration tool box can be used to calibrate the left-right camerasof the binocular stereo vision system. The internal and external parameters of theleft-right cameras can be got. Make the pictures filtered, gray level adjusted,preprocessed the banalization of the picture, so the edge of the laser can be extracted bythe method of canny operator. The continuous contour curve edge of the laser could begot.2）In this paper the stereo matching is completed, the initial matching points are obtained bythe method of feature matching; the error points in the initial matching are eliminated bythe method of polar line geometric constraint. Finally, the matching results are obtained.3）Based on the camera internal and external parameters and the matching results have beenobtained, the three-dimensional reconstruction is completed by the method of visualprinciple of binocular stereo vision, so, the three-dimensional coordinates can beobtained, the minimal radius of curvature of each point in the curve could be calculatedupon the information of the three-dimensional coordinates and the method of the leastsquare circle. An ideal curve could be contoured with the zero curvature points found。The quantitation of surface deviation could be completed via comparing the active curveand the ideal curve.In this paper, the research significance is to detect the quantitative of stamping surfacedeviation, avoid depending on the detector personnel’s subjective evaluation. The method isnot only having high detection speed but also have a high accurate result. If the detectionsystem is applied in the process of industrial production, the detection of surface deviation ofstamping parts could be completed on line.