Research On Color Steel Plate Defect Detection And Intelligent Classification Based On Machine Vision

With the home appliance products consumer market's hot,as one of the raw materials of home appliances,color plates also will be an increasingly wide range of applications.Color plate used in home appliances,the surface quality requirements compared to other uses of steel are more stringent.The topic for this demand designs a color plate surface defect detection and classification system based on machine vision detection technology.The main structure of the system consists of defect image acquisition,defect image processing,defect feature selection and extraction,defect types Pattern Recognition.For the key technical problems in both theory and practice has done in-depth research during the detection,identification and classification process.The main work of this paper is as follows:1.According to the characteristics of the defects to select the appropriate lighting mode.Under the guidance of the principle of the most prominent feature of the defect,image acquisition system uses illumination bright-field and dark-field combination.Self-produced a series of LED diffuse emission light source with continuously adjustable brightness according to the actual needs of the detection system and laboratory conditions.2.Multi thread design scheme is adopted for the detection system software.The software system is composed of image acquisition thread,image processing thread,and defect information database.In order to ensure that the detection system can run in high efficiency,precision,security and stability,all these threads are controlled by a thread controller to achieve a total control of the various threads running.3.Depth study of the image pre-processing algorithms and image segmentation algorithm based on the defect causes and morphology.Through experiment comparison and analysis,median filtering and image gray stretch as an image pre-processing algorithm;Aiming at the characteristics of different defects,in this paper,the integrated use of threshold segmentation,Canny edge detector segmentation,watershed segmentation algorithm and morphological and other methods defects are successfully separated from the background image.4.Study the defect feature selection and extraction method.Principle of feature selection criteria according to pattern recognition,the geometrical features,topological features,gray statistical features,shape features and texture features are used as the input of the classifier in this paper.To improve the effectiveness of feature vector data to classification,using the method of excluding outliers and feature component normalization method to optimize the original feature components to make them have a better classification performance.5.Designed the Back-Propogation neural network classifier model.This classification is capable of oil spots,pitting,scratches,paint bright spots and sand grainwith the five categories of defect recognition accurately after the first training and the second improvement training.Among them,pitting and scratches recognition rate can reach 100%.The experimental results show that the design of color sheet surface detection and classification system for key technology solution is effective and feasible in this paper,and have a certain reference value and inspiration for color plate surface quality inspection.