Machine Vision Based PCB Bare Board Micro Defect Detection System

It is well known that Moore's law in the microelectronics industry means that when the price is unchanged,the number of components that can be accommodated on an integrated circuit doubles every 18-24 months,and the performance doubles.In the context of Industry 4.0 and China's smart manufacturing 2025 strong country strategy,the printed circuit board is the core component of electronic products,and its quality testing has become the key to meeting Moore's Law in the microelectronics industry.Traditional detection methods can no longer meet the high-speed and high-precision requirements in the production process.As a non-contact and non-damage intelligent detection technology,machine vision has become the industry's development trend.In this paper,a bare PCB is used as the research object,and a micro-vision defect detection system based on machine vision is designed.Detect and classify defects such as burr,dimple,broken circuit,short circuit,residual copper,plugging,leaking print,and multiple print in the production process,and meet certain accuracy,recognition rates,certain accuracy and speed requirements.Firstly,the background and significance of the research of this subject are introduced,and the current status of domestic and foreign research on PCB bare board defect detection methods is summarized and analyzed.Secondly,combined with the actual situation of this subject,the overall design scheme of PCB bare board micro defect detection is proposed.It is divided into two modules,one is hardware and structural design module,including the selection of light source,camera selection and overall structure design,etc.The other is a software system module,including image collection,preprocessing,defect classification and identification algorithm and GUI software interface development.This paper focuses on the software system modules of visual inspection,especially the overall process of defect inspection.Grayscale transformation is used to grayscale a color image.By comparing various algorithms,choose logarithmic transformation to increase image contrast.The improved median filtering algorithm is used to remove the salt and pepper noise in the image.Threshold segmentation is performed by the maximum inter-class variance method to obtain a binary image.The target PCB image is registered by the template matching method so that it is in the same spatial position relationship with the standard PCB image.The differential image is obtained by performing a difference operation between the preprocessed target PCB image and the standard PCB image.The morphological algorithm is used to remove small differences that do not constitute defects,and a defect map is obtained.The Sobel operator is used to perform edge detection to obtain the geometric center of the defect area,so as to cut out a defect map with a size of 24 * 24.Using image rendering technology to synthesize new defect images as training label data.The BP neural network algorithm and the convolutional neural network algorithm are used to establish training models as image classifiers to identify various types of defects.Based on the entire set of defect detection algorithms,a GUI software interface is developed to improve the defect detection system.