Research And Implementation Of Intelligent Detection Method For Electronic Components Defect

With the vigorous development of electronic technology,the quality requirements for electronic components are also getting higher and higher.However,due to the limited production technology,when actually producing electronic components,there will be more or less defects on their appearance.The traditional detection method is to use artificial eyes to detect,which has higher requirements for workers.With the increase of labor intensity and other additional factors,the false detection rate and missed detection rate are often high in the process of component defect detection,which is difficult to meet the requirements of modern production.In this thesis,we study the defect detection technology of electronic components’ appearance based on digital images,and uses the method of deep learning to detect various defects of components,which is of great significance in industrial quality inspection.In order to realize the intelligent detection of defects in electronic components,the following aspects have been studied:1.Research on defect detection methods for electronic components.Aiming at the problem that electronic components are difficult to detect with defect sizes,a defect detection method of electronic components based on improved YOLOX is proposed.Soft pooling and efficient channel attention are used to enhance the detection of small target defects.The complete intersection over union loss function is used in the training,so that the average detection accuracy reaches 99.05%.2.Research on lightweight method of target detection model.Aiming at the problems of large parameters and slow detection speed of traditional target detection methods in the intelligent detection of electronic component defects,a new electronic component defect detection method based on lightweight YOLOX is proposed.Through the lightweight processing of the backbone network and the construction of the feature enhancement module,the network parameters have been significantly reduced while ensuring certain accuracy,and the loss function has been further improved.The effectiveness of the experiment was verified on the self built dataset and PCB dataset.The improved model has a detection accuracy of 98.96% and a detection speed of 0.09 seconds per sheet on the self built dataset.3.The realization of electronic component defect detection system.In order to facilitate human-computer interaction and combine algorithms with practice,this thesis designs a defect detection system for electronic components.Conduct defect detection on the input component image and output correct detection information,including defect location,quantity,category,detection time and other relevant information.