Locating and checking of a BGA pin's position using gray level

A machine vision system for SMT-mounting machine applications is usually a two-stage algorithm. It first measures the centroid and rotation angle of the SMD, and then checks each pin's area, position error, and grid coordinate location. In this paper, a set of complete procedures is proposed to locate and check the BGA image. During the locating procedures, we first calculate a threshold of the frame using an iterative threshold algorithm. If an object is found under this threshold, then the pin's area is calculated by the local threshold by using a momentum algorithm. After that, whether this object is a pin or not is decided by its neighbors' relative positions, the approximate rotation angle for finding the outer pins is calculated, and the centroid as well as the rotation angle of a BGA component is calculated by the rectangular least squares algorithm. The checking procedure also measures each pin's area using the momentum algorithm, then it calculates the radius of the moving sum using each pin's area, and finally measures the position error using a moving sum algorithm and judges each pin's type by gray level. The new method uses the gray level statistic information to solve the empty pad's problem and utilizes the symmetrical property of a circle to deal with the shape problem. Lastly, the CRC algorithm is used to check the correspondence between each pin and its pin type. This new method has a high accuracy and a low execution time. It can meet the crucial timing requirement of a high-speed SMT machine through experimental verification.