RTStitch: Real-time Stitching of High-resolution PCB Images
Wenbin Yang, Xianzhong Liu · 2021 IEEE International Conference on Systems, Man, and Cybernetics (SMC) · 2021
In the Automatic Optical Inspection (AOI) system, it is necessary to stitch high-resolution Printed Circuit Board (PCB) partial images in real time to obtain a panoramic image. However, the proportion of the overlapping area of these images is usually relatively small, and there are few feature points in the overlap area of some images. Traditional image stitching algorithms are not effective in processing such images, and it is difficult to meet the requirements of real-time stitching. Therefore, we propose RTStitch, a method specially used for real-time stitching of high-resolution PCB images in AOI systems. We divide the images into two categories. An image with a large change in the pixel gray value of the first column in the overlapping area is the first type of image, otherwise, it is the second type of image. We use different methods to obtain a relatively accurate maximum overlap position for these two types of images. Besides, we propose a new method of selecting template images, then perform template matching to obtain horizontal and vertical errors. We focus on the areas where the gray level changes drastically in the overlapping area to improve the accuracy of the registration. Our experimental results show that for the first type of image, RTStitch improves the registration accuracy by more than 2%, while for the second type of image, the registration accuracy is improved by more than 10%. And more than doubled the stitching speed. Moreover, RTStitch can meet the high-precision and real-time requirements of the AOI system.