Detection and location of mount chip and solder joint based on machine vision.

To realize the automated welding of mounted chips and solve the problems of low efficiency and poor yield rate of manual welding, the detection and localization algorithm of mounted chips and solder joints based on machine vision is proposed. Firstly, to improve the efficiency of the algorithm, we propose a region filtering algorithm with minimized interference to extract image regions of interest (ROI). Secondly, the region-connected domain method is proposed to obtain the set of weld ball centers of gravity and then calculate the radius of the welding ball and the number of empty points. After that, an improved linear fitting method is utilized to calculate the chip body size, average pitch of solder balls, and rotation angle information. Finally, the method proposed in this paper is validated by performing experimental tests on 12 different mounted chips. The experimental results show that the detection accuracy of this method can reach 5 µm for the offset amount and 20 µrad for the angle deviation amount, and the detection time of the unit welding ball is kept within 1.03 ms. Compared with the traditional algorithm, the detection speed is improved while guaranteeing detection accuracy. [ABSTRACT FROM AUTHOR]