Research And Design Of PCBA Defect Detection System Based On Machine Vision

PCBA(Printed Circuit Board Assembly)is a process in which various components are inserted and soldered on a PCB(Printed Circuit Board).With the development of electronic products and the progress of intelligent manufacturing technology,PCBA is developing towards high precision and high density.Due to the production process and human factors,various defect problems will inevitably occur during the production of PCBA,so the defect inspection of PCBA is an essential part.Traditional manual inspection is not only inefficient,but also expensive,and cannot meet the needs of modern industrial production.In recent years,with the development of computer technology,PCBA defect detection methods based on machine vision have been gradually applied.However,due to the subtlety and variety of defective parts of PCBA,it is difficult to distinguish defects on imaging,resulting in a general PCBA defect detection system with poor robustness and unsatisfactory detection rate,so many companies need to customize the development of detection equipment for the characteristics and detection needs of PCBA.In this thesis,we design a PCBA defect detection system with high detection efficiency,low hardware cost and simple configuration for the actual needs of an international well-known manufacturing company,and realize a variety of defect detection functions for power supply fan PCBA,which has high application value.The main work of this thesis is as follows:(1)A PCBA defect detection system based on machine vision is designed and implemented.In-depth analysis of common PCBA defect problems,design of PCBA system architecture,selection of system hardware according to the actual production environment of PCBA and detection needs,selection of suitable industrial cameras,lenses and light sources to form an image acquisition system,design of PCBA defect detection system software interface.Theories and methods related to image processing are studied,including grayscale,image filtering,grayscale transformation,image alignment and geometric transformations theories,completing the acquisition of PCBA images and image pre-processing,completing the extraction of line sequence detection images and THT solder joint detection images,and designing the corresponding detection algorithms for specific detection functions.(2)A line sequence detection algorithm based on LUV(CIE 1976(^*,^*,^*))color space and multilayer perceptron is designed.This thesis studies the color space and machine learning methods,analyzes the effect of PCBA images in different color spaces,and compares the color classification results of different classifiers for line sequence detection images,designs a line sequence detection algorithm based on LUV color space and multilayer perceptron,performs color classification by line sequence detection images,extracts weld line regions,sorts weld line regions according to coordinate relationships,determines whether the weld line welding order is wrong,and completes color classification and sorting of line sequence detection images.(3)A THT(Through Hole Technology)solder joint detection algorithm based on color image segmentation was designed.To study the defect detection problem of THT solder joints with less tin,more tin,and bridging,we designed a THT solder joint detection algorithm based on color image segmentation,using RGB and HSV color spaces and edge detection algorithms to segment THT solder joint images in color,extract defect features of solder joint images and classify them.We also compare deep learning-based methods for THT solder joint detection,and the results show that the color image segmentation-based methods are more suitable for the detection needs proposed by companies and can provide real-time detection solutions in industrial production.The PCBA defect detection system designed in this thesis has been applied to the actual production of the enterprise,and the results of the large-scale inspection on the assembly line prove that the system has the advantages of strong real-time and high inspection efficiency.