Research On Quality Inspection Technology Of Electronic Components Based On Machine Vision

Electronic manufacturing is one of the most rapidly developing industries at present.With the development of electronic components in the direction of miniaturization,the difficulty in quality inspection is greatly increased.On the one hand,the traditional manual detect method is inefficient and has poor accuracy.On the other hand,when using machine vision for quality inspection,there is no quality detection model that can be applied to a variety of electronic components.Aiming at the shape defect of electronic components,this paper explored the machine vision inspection technology applicable to the quality inspection of electronic components and built the shape quality inspection model of electronic components to realize automatic detection of electronic components' quality.The main contents are as follows:1)Analyzed the machine vision application used for quality inspection.According to the problem of poor generalization ability of the quality inspection model,a model of shape's quality inspection of electronic components with generalization capability was proposed.2)Studied the target detection method based on deep learning theory and proposed an electronic component detection and identification method based on YOLO network.Based on the YOLO network structure,simplified the network and fused different granularity features for enhancing the network model's ability to express subtle features.This paper proposed a network structure of target detection and recognition of electronic components.In addition,this paper proposed using the template image of standard components as a benchmark and performs data enhancement through geometric transformations,which greatly reduced the demand for network training for collecting data.3)According to the characteristics of electronic components and images,analysis of common image registration technology,in order to ensure the detection accuracy and stability,selected the Fourier transform image registration method to align images.Firstly coarse estimated in the frequency domain and then fined localization in the neighborhood achieved the accuracy of sub-pixel registration.Analyzed image segmentation methods and extracted electronic components with iterative method.Components' defects were extracted by comparing with the standard component images.Finally this paper used the principle of small probability events to determine the defect area threshold for determining the quality of electronic components.4)The experimental verification of the elastic stranded pins in the aerospace electronic connector was carried out.The results showed that the proposed method for detecting the quality of electronic components can effectively detect the shape quality defects of the sample components.The test accuracy rate was 97.87%.Finally,a software system was completed with web application technology.