Study On The Visual Detection Method And Experiment Table For Dashboard Framework Of A Car

The competitiveness of cars largely depends on its quality. Auto body is an important part of car products, its weight and cost accounting for 40 to 60 percent of the entire cars. A large number of case indicated that the problem of autobody quality rooted in the manufacturing stage of assembly. The accuracy of assembly directly affect the whole quality of the car.Dashboard framework is one of the autobody assembly. it fits to kinds of apparatus Because the dashboard framework is a complex and welding part, it may cause some manufacture deviation inevitably. If the unqualified product transfered to the general assembly line of car, it may affect the accuracy of kinds of apparatus, therefore, the quality of dashboard framework should be under control.Visual detection is widely used in industrial measurement because of the advantages of high speed,high accuracy,flexibility and contactless. In this paper, I made a visual detection scheme according to three failure modes (leak welding, position deviation, radius dimension deviation) of BORA dashboard framework.Through the study on the key technology of visual detection method, I established the visual detection system of BORA dashboard framework.Finally,I test and verify the visual detection system through experiment.In the aspect of detection scheme design, I used three groups of stereo binocular visual detection system on detecting 8 selves of dashboard framework.This task is to evaluate 3 failure mode of dashboard framework. Specifically, to judge leakage by the existence of self edge, to judge position deviation by the comparison of the detection and ideal data of the center of the mounting holes. After scheme designing, I built a visual detection experiment table of this dashboard framework, and developed a positioning device for the automatic positioning equipment in the detection line. In the aspect of camera calibration, I used a vertical shifting checkerboard as target, and built a three-layer adaptive neural network to calibrate the camera, finally, the comprehensive calibration error is 0.1577 mm.In the designing of edge detection method, I made a series of pretreatment by eliminating background, denoising the image,enhancing the edge and extracting edge of the image by canny operator. In the designing of stereo matching, it is completed by series of progress including feature matching, extreme line restraining, regional matching and uniqueness restraining, finally built the one-to-one relationship.In the aspect of hole center calculating, I solved the elliptic equations of two hole edge first, and then calculated the center coordinates of each elliptic, finally,calculated the hole center coordinates by neural network. In the aspect of hole radius calculating, I calculated the space coordinates of each matching points first, and then calculated the average value of the distance between space coordinates and hole center as radius. Finally,through the error analysis, I got the visual detection system error is less than 0.1mm, the result meet the dashboard framework detection requirements.If the visual detection method and positioning device used in the actual on-line detection of BORA dashboard framework, the significance of the study lies in the effective control of BORA dashboard framework quality through visual detection accurately and rapidly.