Research On The Key Technology Of Measuring The Shape Of Automobile Complex Components

Generally speaking,the large-scale surface three-dimensional shape measurement system often requires its own high online flexibility and effectiveness.However,in the process of using some commercial measurement systems which have come out at present,they usually need to paste points or to splice point clouds globally based on robot base coordinate system,which greatly reduces the measurement accuracy and efficiency,and affects the complete expression of 3D information of products.Therefore,in view of the above problems,this paper attempts to build an online flexible measurement system,which combines the shape sensor with the robot,uses the IGPS to track and locate the shape sensor globally,and realizes the efficient registration of point cloud data under non stick point measurement.The main research work of this paper is as follows:Firstly,the mathematical model of the measurement system is constructed by using ATOS scanner,Yaskawa HP20,robodk off-line programming software and IGPS positioning and tracking system,and an adaptive measurement method is proposed based on CAD digital analog information.Compared with the traditional equal step measurement method,this method has higher measurement accuracy and efficiency.In addition,based on the mathematical model of the measurement system,a hand eye calibration method integrating 5 feature points is proposed,and the transformation relationship between the coordinate systems is theoretically deduced.Secondly,the forward and inverse kinematics algorithm of HP20 robot is deduced and simulated.At the same time,several commonly used trajectory planning methods are analyzed,and the simulation results in joint space and Cartesian space are compared by MATLAB software to verify the stability and practicability of the robot motion process.Then,the common feature point detection operators are analyzed and compared,and an Improved SIFT algorithm based on octree KD index structure is proposed.By using the octree KD structure index,this method can speed up the location of the nearest neighbor points and shorten the search time of the algorithm.At the same time,the improved RANSAC algorithm and the cosine angle of normal vector are used to double screen the feature points,which improves the accuracy of registration.Finally,through the use of Visual Studio software and the third-party tool open-source point cloud library PCL,the automatic registration of the door and the front cover and other instances is realized.Experiments show that:(1)every registration error of point cloud data is controlled within plus or minus 0.1mm;(2)the registration time is 30% less than the original algorithm.