Research For PCB Online Detection System Based On Machine Vision

In recently years,the direction of electronics industry turns to the development ofminiaturization and high integration. Thus it puts forward the design requirements ofhigh-density multilayer for the PCB manufacturing process which is an importantsupport of the electronics industry. The quality of PCB and post-welding technologydirectly affect the quality of the product. As the high density of modern PCB wiring,numerous number of components and their small size, it is more easily to bring varietyerrors. Contact measurement and manual inspection not only unable to meet therequirements of the detection precision, but also can not meet the requirements oftesting speed. The introduction of machine vision method provides a new way to solvesuch problems.Machine vision technology is a non-contact detection, and the PCB imageinformation is obtained by the CCD camera. According to the image pretreatment, wecan distinguish the PCB to find whether it exist defects and the number and location ofdefects based on the image content analysis. Detection speed and high accuracy are theadvantages of this method. In this paper, we try to design a PCB defect detection systembased on the research of machine vision detection algorithm.First of all, this paper analyzes the current research status based on machine visioninspection system, then given two sets of hardware configuration program with theproduct information of collecting the light source, camera and image capture cardshardware. Based on the theory of image processing and pattern recognition, this studyalso focus on the common defects of PCB and several detection methods proposed bythe previous researches, complete treatment programs were put forward by comparingtheir advantages and disadvantages.In the links of enhanced image, it uses the methods of histogram equalization andmedian filtering so as to improve the image quality and suppress the noise. In the linksof image segmentation, we analyze and compare several classic image segmentationmethods, then choose the bimodal law as image segmentation method based on the testresults. In the links of PCB defect recognition, we use the wire tracking algorithm toidentify the wire circuit and pits other defects, then use the8-connected method toidentify the short-circuit and convex defects. The usage of regional statistics is toidentify the stain. For the detect component counter and welding leakage, we use theoptimization template to match the algorithm. The experiment results show that thedefect detection algorithm can effectively identify the defect.In this thesis, PCB defect recognition system was developed by using the Matlabsoftware, the programs achieve the goal that control the light source, image acquisition,image preprocessing, PCB bare board circuit, short circuit, stains, lack of weld,anti-fitted detection and also achieve the function of the statistical analysis of test results.Operation interface of the system is concise and clear, with the experiment testing, the system possesses the high reliability that it can accurately identify the minor error.