Research On Robotic In-process Inspection Technology And System Of Body-in-white Assembly Features

Body-In-White (BIW) is an important part of the automobile. Itsmanufacturing quality and precision directly affects the matching ofautomobile’s interior and exterior decoration and determines automobile’sperformance. Strict control of the BIW dimensions is a key point toensure the manufacturing quality and production efficiency in automobileindustry. Therefore, it is of great significance in engineering to monitorBIW assembly dimensions in process by measuring the position andorientation of various assembly features in different parts of the weldedBIW. The existing technologies home and abroad can be divided into twotypes, contact inspection and non-contact inspection. Among them, thenon-contact robotic inspection based on machine vision is the trend ofresearch because of its high efficiency, flexibility and moderate accuracy.Nevertheless, there are still deficiencies in ambiance light control, sensorpose control or robot motion errors in such technology. As a result,research of new technologies is necessary to overcome those deficiencies.Technology of BIW in-process inspection based on laser scanning isresearched and a solution applying swinging laser stripe monocular visionsensor on robot system for BIW in-process inspection is proposed: Thesensor is carried to the best measuring position by robot; The turntableinside drive the sensor swinging and trigger the camera to capture a set oflaser scanning images of BIW assembly features; The inspection of BIWassembly features is realized automatically, stably and accurately byimage laser point extraction,3D laser point cloud integration and featurepose parameters calculation.The swinging laser stripe monocular vision sensor is researched, including optimized design of its optical structure and turntable assemblyand reasonable selection of optical components and mechanical parts.Free pose calibration method for sensor’s optical parameters andefficient calibration method for sensor’s turntable are proposed in order toovercome the deficiencies in the existing methods and meet the need ofswinging motion calibration. One planar target is shared. Projectiveinvariance of a set of vector products’ orientations is analyzed andcross-ratio invariance is applied to realize the senor calibration in freepose. Ideal results are obtained in calibration experiments.Vector computing for BIW feature images is researched. Methods oflaser point cloud extraction and pose parameters calculation aiming at avariety of BIW assembly features are proposed, including hole, threadedhole, complex hole, cylinder and edge. The extraction starts from imagecenter and algorithms of image correctness and regional connectivityjudgment are added. Pose calculation in reduced dimension is realized byestablishment of local coordinate system. Methods are verified byexperiments.Prototype was developed and in-process inspection experimentswere implemented in a BIW manufacturing site. The absolute accuracy,repeatability and inspection efficiency were evaluated and ideal resultswere obtained. Sources of measurement errors were analyzed anddirections of future research were discussed.