Key Technology Research For Non-Contact Inspection Of Gaps On The Car Body

One of the most difficult tasks in the later stages of automobile assembly is the dimen-sional inspection of the gaps between the car body and the various panels fitted on it (doors、motor-hood，etc).Because the gaps not only affects the vehicle’s wind-proof and rain-proofcapability，but also closely related to the vehicle’s vibration and noise’ size and the per-formance. The width and the flush of the gap is one of the quality requirements for the vehi-cle’s appearance, so the inspection of produced cars is necessary.Now，the gap-measuring task in the automobile industry has been automated at a veryslow pace and it is still performed by humans. Using humans for the gap inspection requiresconsiderable manpower that significantly increases the costs and the time for inspection，andthe method provides low precision，subjective results and many errors are disregarded due totime limitations.Visual detection is widely used in industrial production because of the advantages ofhigh speed、high accuracy、flexibility and contact less. So，in this paper，we introduce asystem for gap inspection that employs computer vision，using this method we can achievethe purpose of high-speed、high-precision and not-contact detection.In this paper，we introduce a system for gap inspection based on stereo vision，usingtwo cameras for the gap inspection and choose the gaps between the doors as an experimentexample. The paper’s focal points are extracting contour、stereo matching and the3D re-construction and so on，in addition，we introduce the way for gap on line detection. Thecorresponding work are given as follows：1. First，I used the Calibration Toolbox for camera calibration，got the internal pa-rameters and the external parameters of two cameras；Process the images from the camerasby filtering and grayscale adjustment and other pre-processing. I got the gap’s smooth andcontinuous contour by using the Snake model based on the level set evolution without re-initialization.2. I designed the method for stereo matching，first，using the feature matching basedon the feature points of the gap’s contour，then searching the left images’ matching points inthe right image based on the epipolar line constraint to get the initial matching points，atlast，excluded the mismatching pints by using the disparity continuity constraint and thesequential consistency constraint，then I got the final matching results.3. Based on the cameras’ internal parameters and the external parameters and thematching points，I got the three-dimensional coordinates of the points on the gap’s contour inthe space，based on this，I can do the task as fitting the gap line and the two planes composedthe gap line，then got the width and flush of the gap，and the result has an accuracy less than0.1mm.The purpose of this paper is improving the degree of automation for gap inspection andthe inspection accuracy and the speed，and using this inspection system in the industrialprocesses in the future，achieving the aim of automated gap inspection.