An Optical Inspection System for the Solder Balls of BGA using Support Vector Machine Classification

Abstract

In this paper, the Support Vector Machine (SVM) is utilized to classify the defective solder ball of ball grid array (BGA) so as to promote the accuracy and efficiency of the optical inspection. The terms of inspections include the ball displacement, the ball over scale, the ball undersize, the ball absence, and ball bridging. There are two step of the SVM processing: the training step and the real testing step. The SVM is more efficiency because of the calculation in testing step is no relation to the number of the input vector. As this point, SVM is fit to build up the high dimension input recognize system such as BGA inspection. From the real testing results, the Support Vector Machine is proved to properly identify and classify the shape defects. The two dimensional BGA optical inspecting system is implemented by Visual Basic as the developing tool incorporated with the Halcon's function which is the database of the image processing on Windows operation system. For the development of the processing procedure of the automatic optical inspecting system with the designed lighting system, the precise geometrical information of the solder ball is evaluated by the sub-pixel method to identify the shape of solder ball and its location under the BGA board is offset and rotated at any angle.