Multi-type BGA Chip Visual Recognition Method Using Line Based Clustering

摘要

A multi-type BGA chip visual recognition method using line based clustering to solve the problems of BGA inspection algorithms which have limited applicability, low flexibility caused by template matching, poor robustness to illumination and interferences and high time complexity. The method is based on: dynamic threshold segmentation of original image, morphological and connected component labelling; BGA extraction based on grayscale connected component and building information list for all complete grayscale BGA solder balls; Building of BGA solder ball label image; local analysis to determine the coarse orientation angle of equivalent BGA array; determining equivalent BGA solder ball clusters for each row and each column and the equivalent BGA solder ball cluster for the boundary by line based clustering; obtaining an accurate orientation angle and center position of BGA chip in the original image by line fitting of boundary BGA solder ball clusters; and extracting different standard parameters of BGA chips.