Research On Online Feature Extraction Algorithms For BIW Assembly Quality Evaluation Based On Line-Structured Laser Scanning

As the carrier of other parts of automobile,body-in-white(BIW)is a core element in vehicle manufacturing.Its manufacturing quality,especially dimension quality,largely determines vehicle's performance.And error generated in the welding process will directly be reflected in the relative position of assembly features of BIW sheet metal stamping parts.Therefore,it is of great significance in engineering to strengthen the precision and adaptability of domestic BIW online detection systems which can effectively realize the real-time monitoring of BIW manufacturing dimensions,through the study of BIW assembly feature extraction algorithms.Based on the robot online detection system,line-structured laser scanning method is used to extract the position and attitude parameters of BIW assembly features.In engineering application,many factors put forward higher requirements in online feature extraction algorithms for BIW assembly quality evaluation,including manufacturing error,interference of surface attachments,variation of ambient light,system relocation error,restriction of production rhythm and line laser measurement,interval scanning measurement of line-structured laser and so on.Therefore,it is necessary to do targeted research to solve these problems.Characteristics of laser curve images were analyzed,laser contours in the extreme value method were used to narrow the calculation range,and then the gray centroid method was improved to achieve rapid and stable laser point cloud extraction.According to laser contour points of hole and slot and strong noise interference of cylindrical laser point cloud,further research was did to improve the efficiency and stability of numerical results.According to geometrical characteristics of hole,slot and edge,basic pattern detection was used for fast segmentation of features,and then mutual position relation was used to find methods for parameter extraction.Secondly,aiming at the problems in practical engineering application,effective solutions were put forward to improve robustness of algorithms,with characteristics of stamping forming and theoretical values of numerical model.Through the analysis of curves in single laser image of cylindrical features,fast segmentation of cylindrical and planar point clouds was did through line detection and the ratio of point number.Aiming at cylinder,especially stud,a vector calculation method of central axis based on statistical analysis of data was proposed.In addition,cylindrical models,data in engineering application and different measurement postures were applied to analyze the accuracy,stability and influencing factors of this method.Corresponding solutions were also available.Based on tests at a BIW online inspection station in an automobile production base,a large number of measured data was obtained to evaluate the absolute accuracy,repeatability and efficiency of algorithms,and error sources were analyzed.These algorithms were applied to the online detection system and put into practical production,and ideal engineering application effect was obtained.