Research On Assembly Defect Detection Of PCB Surface Components Based On Machine Vision

With the development of smart grid in China,higher requirements are put forward for the manufacturing of smart grid equipment.Printed circuit board(PCB)is one of the main components of all kinds of power grid equipment.The quality of PCB production directly affects the reliability of smart grid equipment,so the defects detection of PCB production line is very important.At present,the detection technology based on machine vision has been widely used in PCB defects detection,but most detection systems are designed for PCB bare board circuit detection or solder joint detection,lack of assembling defects detection for PCB surface components.In order to improve the quality of PCB production,based on the full investigation of various domestic and foreign research results of PCB detection technology,aiming at the assembling defects of PCB surface components,the thesis carried out in-depth research on PCB image mosaic,PCB surface component image positioning and component assembling defects detection.The following is the main research content:The overall framework of PCB surface component assembling defects detection system is designed.In the hardware part,the appropriate industrial camera,lens,light source and other hardware equipment are selected;in the software part,the algorithm research of image mosaic,component positioning and assembling defects detection is the focus of the thesis.The PCB image mosaic method is studied.An image mosaic method based on improved surf algorithm is proposed,by introducing the feature point limiting function to accelerate the feature point extraction speed,the efficiency of PCB image stitching is effectively improved while ensuring the accuracy of stitching.The positioning method of PCB surface components is studied.For PCB chip positioning,the chip positioning algorithm based on RGB color features is studied.For electrolytic capacitor positioning,an improved Hough gradient method is proposed based on the study of Hough circle transformation method and Hough gradient method,which improves the positioning accuracy of electrolytic capacitor by introducing error extremum.For SMD components positioning,the positioning algorithm of SMD components based on Hough line transform is studied.Experiments are carried out to verify the effectiveness of each positioning algorithm.The assembling defects detection method of PCB surface components is studied.The image correction method based on mark point positioning is studied to correct the image to be tested.For the component existence detection,three existence detection algorithms based on perceptual hash,H-S two-dimensional histogram and template matching are compared.For component polarity detection,the electrolytic capacitor polarity detection algorithm based on OTSU and pixel statistics and diode polarity detection algorithm based on quadrant segmentation are studied.For the error detection of chip resistors,the character recognition algorithm based on convolutional neural network is studied,and the network is designed to train the Arabic numeral samples of chip resistors.Experiments are carried out to verify the effectiveness of each defects detection algorithm.