| Miniaturization of electronic devices and integrated circuits(IC) brings the higher demands of line-width denseness and multi-layer to PCB making. Multi-layer PCB is made by compacting several layers with epoxy resin, after each of which being etched and detected correctly. If there is problem in any layer of the PCB, the entire multi-layer PCB will collapse. Therefore, the high-precision measurements of the PCB line-width, angle and welding board diameter are vital guarantees to manufacture a high-quality product. In the past, Hundreds-X lens and quadratic were usually applied. Hundreds-X lens depends highly on workers` experiences, with a typical accuracy of 0.1mm and slow-speed. Quadratic method, with the accuracy of 0.01mm, is measured manually, therefore operated inconveniently. Using image-based measurement technology, our group developed a high-precision measurement system, the typical accuracy of which is higher than 3um. This system is applied to many PCB manufacturers and achieved favorable social and economic benefits. Based on the existed system hardware, to advance system measurement accuracy, this research mainly focused on the following studies:(1) Study of the method of image nonlinear correction. Because of the inherent property of optical system, the digital image which is acquired by CCD imaging sensor has non-linear distortion, so the measuring precision is affected. The distorted image is corrected in two aspects, position and gray. First, build distortion correction model, extract the ideal coordinates and distorted coordinates of feature points from a selected typical distorted image, solve the distortion coefficients using the optimal approximate solution under criterion of the smallest error sum of squares. Then, bring the distortion coefficients into distortion model to implement position correction of the whole distortion image. Finally, adopt bilinear interpolation method to implement gray correction, as the position correction results are possibly not integer. The experiments show that distortion correction obtains favorable results.(2) Study of the method of sub-pixel edge points abstraction. First, the image is preprocessed through image correction, median filter , SUSAN pixel edge detection and image segmentation, which becomes binary image. Then, a region of interest is selected and edge points of the region are obtained, the sub-pixel edge location of edge points is implemented by OFMM operator. Finally, the singular sub-pixel edge points are eliminated through least square line fitting, real edge points are reserved for subsequent high-precision measurement.(3) Study of the method of high-precision measurement of line-width and angle between lines. First, the line equations are got separately by line fitting of effective sub-pixel edge points. Then, line-width is calculated by using the average distance of point to straight line, angle is calculated by using the average angle between lines.This paper based on PCB measurement system developed by this research team has advanced the measurement precision of the entire PCB angle and line-width, by means of the study of image nonlinear correction, sub-pixel edge points abstraction, line-width measurement and least square line fitting. |