Research On The Detection System Of Weld Defects Of Silicon Steel Sheet Based On Machine Vision

In this world where computer technology is constantly updated,the application of machine vision in the industrial field will become more and more extensive in the next few years,especially the manual inspection process that consumes a lot of manpower.The use of machine vision in the industrial field can improve product quality and improve inspection.Speed ??etc.In the silicon steel sheet weld defect detection process,the manual detection method is easily affected by the fatigue and physical and mental health of the inspector.At the same time,the manual detection is time-consuming,and the detection standards cannot be unified every time.Therefore,an automated detection system is used instead of manual labor.Detection becomes a trend.This paper designs a cost-effective system for silicon steel sheet weld defect detection.The system integrates multiple modules such as motion control,machine vision,image processing,and software algorithms,which can meet the needs of enterprises for automatic silicon steel sheet weld detection.demand.The main tasks completed in this paper are as follows:First,the research background of the silicon steel sheet weld defect detection system and the significance of its application in practical engineering are summarized,and the current progress and future trends in the field of silicon steel sheet weld defect detection technology at home and abroad are studied.Secondly,the process of silicon steel sheet weld defect detection is studied,and the overall technical scheme of the silicon steel sheet weld defect detection system is designed according to the main technical indicators and detection process.Thirdly,a scheme for detecting weld defects of silicon steel sheet based on HOG+SVM is designed,and four commonly used feature extraction algorithms of LDA,PCA,2DPCA and HOG algorithm are introduced,and the four feature extraction algorithms are compared for the detection of weld defects of silicon steel sheet.Average recognition accuracy,select the HOG algorithm with the highest accuracy to extract the features of weld defects of silicon steel sheet,introduce the principle of SVM classifier and model construction,use SVM to classify weld defects,and penalty coefficient and Gaussian kernel function parameters Made adjustments.Fourth,the design of the weld defect detection scheme of silicon steel sheet based on Faster-RCNN neural network,the structure of Faster-RCNN neural network,and the training methods and procedures of Faster-RCNN are studied.Fifth,the Faster-RCNN neural network is optimized and improved,and the Res Net＿50deep residual network is used to replace the original VGG＿16 network model as a pre-trained network model for the basic training of Faster-RCNN.Finally,the results of the two defect detection schemes of HOG+SVM and Faster-RCNN neural network are compared and analyzed,and the Faster-RCNN neural network with higher recognition rate is finally used as the detection method to design the silicon steel sheet weld defect detection system.