Improvement of FLIP CHIP detection and localization based on traditional image algorithm

Abstract

Surface mount technology (SMT) in integrated circuits, is one of the pillars of electronic information industry. Its degree of development is also one of the core indicators to measure the level of science and technology of a country, which affects process of social information. In recent years, chip technology has improved, and surface mount technology has also achieved remarkable results. But there are still two problems, one is for high-density integrated chips (such as BGA, FLIP CHIP, etc.), the recognition of pass rate is low by the light, chip quality, and other external factors such as more significant impact. The other is traditional template matching algorithm used for positioning, which takes a long time and leads to low production placement speed. In this paper, we take FLIP CHIP as an example and propose a regional screening algorithm to reduce interference, which includes area screening and roundness screening to extract the most valuable part of image information. The regional connected domain method is proposed to obtain the center of gravity set of weld balls and the circle fitting function in OpenCV is used to calculate weld ball radius. The pre-processed image is taken to improve linear fitting method to calculate the chip body size, the average pitch of solder ball, the rotation angle, and the coordinates of the center point of the chip. The remapping search algorithm is used to calculate the number of empty spaces. Then, the rotation angle under the obtained image coordinate system and the center point coordinates of chip are converted to the physical coordinate system at the time of positioning. Then, the offset is output by subtracting from the size of acquired ROI. It is more advantageous than the method of template matching to improve detection speed while ensuring accuracy.