On-line Defect Detection Of Online Thin Steel Sheets Based On Machine Vision System Design

With the development of machine vision technology,image-based inspection technology has always been an important branch in the field of image processing.In recent years,due to the large increase in the processing volume of thin steel plate parts,image-based steel plate surface defect detection research has become numerous.Due to the variety of defects on the surface of steel plates,the traditional detection technology has been unable to meet people's needs for their accuracy and efficiency.Machine vision-based surface defect detection methods have the characteristics of high accuracy,relatively intelligent,and fast speed,which is a major trend in surface defect detection.With the development of machine vision technology,image-based inspection technology has been continuously used in industrial production.In recent years,due to the large increase in the processing volume of thin steel plate parts,image-based surface defect detection research has become a hot topic in many fields.Due to the many types of defects on the surface of steel plates,traditional detection techniques have been unable to meet people's needs for their accuracy and efficiency.The accuracy and low efficiency of traditional detection methods can no longer satisfy people and society.Machine vision-based surface defect detection methods are characterized by high accuracy,fast processing speed,and intelligent processing,which is a major trend in surface defect detection.This paper takes the surface defects of online thin steel plates as the research object and uses traditional machine vision inspection methods to study and design the surface defect detection systems of online thin steel plates.The main tasks are as follows:(1)Aiming at the problem that the traditional median filtering and Gaussian filtering algorithms are insufficient to retain the image edge details when filtering Gaussian noise and salt and pepper noise,a partial differential median filteringalgorithm is proposed,and in order to increase the defect target in the steel plate image Contrast with the background,image enhancement was performed on the steel plate image.By comparing the histogram equalization and the segmented adaptive gamma function conversion image enhancement algorithm,the segmented adaptive gamma conversion function was selected for the steel plate defect image.Image enhancement.(2)Experimental verification shows that traditional edge detection and threshold segmentation techniques cannot segment the surface defect features of steel plates.Finally,an improved threshold segmentation algorithm based on pixel search is proposed,and it is found that the algorithm perfectly solves edge detection.Coordination of texture detail preservation and resistance to noise.(3)Based on the traditional HOG feature extraction,information entropy weighting and PCA feature dimension reduction are introduced to improve the traditional HOG feature extraction algorithm.Experimental verification shows that the algorithm has significantly improved its effectiveness and applicability.Finally,it uses support The vector machine trains and identifies the extracted defect feature information.(4)Research on the surface defect detection system of the thin steel plate,select the hardware facilities required by the detection system,and finally display the defect detection classification results on the interactive interface.