Design And Implementation Of A Positioning Scheme Based On Binocular Stereo Vision For Chip Mounting Slots

The electromagnetic compatibility test of chips is an important part of the chip packaging test assembly line.When testing,a chip need to be plugged and unplugged into the chip mounting slot on a test board.Considering this application scenario,this paper focuses on how to achieve precise positioning of the chip mounting slot based on binocular stereo vision technology so as to design a set of chip automatic plugging robot system.The specific work of the thesis is as follows:(1)A hardware system prototype of binocular stereo vision system was built.Based on the calibration method of Zhang Zhengyou,the parameters of binocular system were calibrated,and the calibration result with an error of 0.09 pixels was obtained.It paved the way for subsequent image correction and target spatial positioning.(2)The target detection algorithm of chip slot area was designed and implemented.The yolov5,which takes into account both detection accuracy and speed,was selected to realize the detection and recognition algorithm of slot area based on experimental comparisons.By the algorithm,the system can obtain the coordinates of the upper left corner and the lower right corner of the slot area,and achieve positioning of the target area.(3)An optimized stereo matching algorithm was put by both improving the census transform algorithm and fusing it with absolute difference cost.It helps to reduce the matching algorithm's dependence on the center pixel of the window,and enhance the algorithm' noise resistance by improving the census transform algorithm.Moreover,it can alleviate the ambiguity of the matching algorithm on repeated textures by combining the improved transformation algorithm with the single-pixel-based absolute difference algorithm for normalized cost fusion.(4)A solution to 3D reconstruction of the center of the slot area using a back projection matrix was proposed.Experiments showed that the error of the 3D reconstruction was controlled at about 1mm.In addition,on the basis of the algorithms above,the software functional modules were designed,as well as a set of human-computer interaction system for binocular vision positioning of chip mounting slots,which included image acquisition,camera calibration,target detection,stereo matching,and target positioning,was realized.The system is convenient to put into engineering application.The work of this paper gave some technique accumulations for the subsequent implementation of a chip plug-in robot automation system,but also provided a reference example for solving other spatial positioning problems.