In view of the position accuracy requirement of industrial transmission chain in latch assembly process, a low cost, low computational power and high precision latch automatic adjustment system based on machine vision has thus been proposed. In the image acquisition stage, the interference background is effectively and quickly removed by adding frame number acquisition constraints to the background difference algorithm. In the position acquisition stage, the particle swarm algorithm is introduced into Hough transform to speed up the pin edge feature positioning, with the angular position of the pin calculated by combining the minimum enclosing moment. Compared with other common matching algorithms, the results show that the improved method can not only improve the computational speed but also obtain a higher computational accuracy.