Research On The Algorithm Of Chip Recognition And Location Based On The Template Matching

In tradition,Chip mounting and alignment process rely on manual work.Accuracy requirement is improved at micron level,and it needs massive manpower and accuracy is no guarantee,so it is of great significance for the study of rapid chip identification and localization.The chip is interference by complex background,such as occlusion,unstable illumination and deformation,and there are some difficulties of low identification efficiency and stability.This paper makes an intensive study for rapid chip identification and localization above problems,and specifically carries out the following work:(1)For the requirements of chip identification and localization,we propose a chip real-time and stable localization scheme of template matching by gradient feature of edge.We focus on the study of feature selection,similarity measurement scheme,and edge feature extraction algorithm.(2)We propose an improved edge algorithm of gradient judgment of object boundary,where the interference of the same amplitude of the adjacent anchors in the gradient direction is eliminated,and the sub-pixel edge is obtained.Then pin profiles are extracted by the eight-connected domain and watershed algorithm.The center line of pin is fitted to align the pose of center point of chip,and the precise position and posture of chip is realized.(3)This paper studies the identification and localization of template matching based on gradient features of pin.We propose improved cosine similarity calculation that is combined the method of fast Fourier transform and pyramid model to improve the matching speed.Then,the improved genetic algorithm is proposed to accelerate the optimization search process.The matching strategy of coarse-to-fine is realized,and the efficiency of the sub-pixel accurate identification and positioning on object image is improved.In addition,for flexible deformation of object,we proposed a deformable model to accurately locate the object.(4)The software about identification and localization by template matching is developed under the Python + Open CV platform.The pose accuracy of the template matching is tested on the five-axis motion platform.The experimental results show that proposed algorithm is fast,high accuracy and stability.The matching algorithm takes 800 ms at images of 2592×1944 resolution,and the object can quickly be identified under environment of unstable light condition and occlusion,with an accuracy of 97.5%.The rotation angle and position matching error was0.035°,2.7μm,respectively.