Research On Surface Quality Inspection Technology Of Rubber Hose Mandrel Based On Machine Vision And Neural Network

The visual defect detection based on neural network is a combination of machine vision image processing technology and neural network data mining algorithm.In this paper,the machine vision technology is used to detect the surface defects of the mandrel of the rubber hose,and the image processing,ROI segmentation,boundary extraction and threshold segmentation are adopted to identify the defects of the mandrel,which realize the identification,quantitative analysis and get the output of the defect data.On the basis of obtaining the defect data,the method of evaluating the mandrel surface quality is obtained by the neural network data mining algorithm and building classification model.This method can improve the status of artificial detection and has important research value and development foreground.At first,the paper introduces the production status of the automobile industry and the automobile rubber hose,and summarizes the background,the development status and the future trend of the machine vision technology.Then combining machine learning classification and algorithm,the key technology and principle of machine learning process are studied in depth: This paper introduces the basic concept,algorithm,principle and research actuality of machine learning,introduces the principle of neural network algorithm and error correcting learning algorithm,and puts forward the method of mining defect data by using neural network model.Combining with the process of data processing,the algorithm and process of machine vision defect recognition are studied,a recognition algorithm of the surface defects of the mandrel is proposed,and the quantitative analysis of different kinds of defects on the mandrel surface is realized on the basis of satisfying the recognition accuracy rate.Finally,according to the actual problem of the mandrel surface quality assessment,the data mining platform Rapid Miner is used to complete the defect data mining,and the neural network model is constructed to predict the surface quality of the mandrel.The modeling and output of the evaluation standard are realized.In this paper,a machine vision detection system is used to image the mandrel,and the image is used to identify the defects.Then,the data of the defect outputted by the system is mined,the neural network classifier model is constructed,and the modeling results are analyzed and contrasted by means of cross-validation,which verifies the feasibility of the model’s accuracy and surface quality evaluation method.