Pcb Photoelectric Detection System Appearance Standard Testing The Establishment Of Databases And Research

Along with the rapid development of the PCB industry, PCB defect inspection has been paid more and more attention by PCB manufacturers. Compared with the traditional manual testing method, the PCB visual inspection system is more efficient and more accurate to determine the quality of PCB.The detection principle of PCB visual inspection system is the comparison between to-be-detected image of the PCB and PCB standard detection template, finding out the differences between the two, and determining whether the differences are defects. Therefore the precision of stand detection template is directly related to the results of defect detection, which is also a prerequisite of the system’s subsequent image processing. Creation of standard information detection library contains image pre-processing, optical positioning hole center positioning, creation of standard detection template, etc. So it is necessary to do a further research.In the image preprocessing step, two kinds of image filtering method including mean value and mid-value are compared and analyzed. Gray-scale image edge detection and color image edge detection are compared. Considering the noise effect and optical center positioning hole of arc might be damaged, an improved random Hough transform algorithms for center positioning is chosen. The experimental results show that center positioning can be accurately realized by this algorithm.Establishing the standard detection template is the core content of the standard detection information library establishment, two methods are analyzed in this dissertation. In constructing the standard template based on the standard image of Bayesian decision theory, the advantages and disadvantages of three kinds of color space including RGB, HSV, CIELab are compared, and the CIELab color space is selected to classify Bayesian color. The method of constructing standard template based on analytical Gerber file is introduced, and the analytic process of Gerber file is elaborated. Experimental results of the two methods are compared, and their respective limitations are analyzed. Finally, the common defects on the PCB as well as the recognition method are introduced.This dissertation based on the2009Guangdong’s key areas of breakthrough projects "intelligent PCB final visual inspection machine", which is a cooperative research and development project between our laboratory and Guangdong Zhengye company. To our knowledge, the method of using Bayesian decision theory in the CIELab color space to build a standard image has not been reported in other similar products.