The Printed Circuit Board Defect Detection Based On Machine Vision

The Printed Circuit Board (Printed Circuit Board, PCB) defect detection basedon machine vision is the ingretation technology of image acquisition, imagetransmission and image processing, it is the future development trend of non-contactdefect detection technology and has many advantages, such as high intelligent level,defect detection coverage and error detection rate. Printed Circuit Board is the basicmaterial of electronic communication products, whose quality affects the overallquality of the electronic communications products and system stability. Therefore, thedefect detection method carried out on the PCB is critical to improve the quality ofthe products. Because of its high intelligent level, defect detection coverage and errordetection rate, it has been widely used in defect inspection of PCB.This project cooperates with Zhejiang OVI Technology, designing and producingthe embedded FPGA hardware module and taking the advantage of parallelcomputing and assembly line processing of FPGA processor to solve the complexlogic operations. It uses upper machine to implement part of the image processingalgorithm, thus improving the defects detection speed and recognition accuracy ofPCB-AOI products, which has the very good practical value.The main contents of this paper are as follows:1)It designs the PCB image transmission collection and processing system,including the image acquisition module based on Camera Link interface, the FPGAimage processing module and the pixel coordinates data transmission module basedon Ethernet MAC (Media Access Control) protocol layer.2)It proposes using Verilog hardware description language to implement thealgorithm of contrast enhancement based on gamma transformation of copper andnon-copper area in PCB image on the FPGA image processing module, it alsoeliminates the noise of median filtering algorithm and preprocesses the scanned PCBimages.3)This paper focuses on the design and implementation of the imagesegmentation algorithms based on double threshold, gradient threshold and gaussLaplace transformation, segments the copper and non-copper region in PCB image and extracts the sub pixel coordinates of the binary image after segmentation.This paper achieves the real-time image processing on the PCB based on theparallel processing and pipeline technology of FPGA processor, reducing the pressureon the PC’s operation, which is of high practical value.