Design Of On-line Detection Of Edge Defects Of Cold Rolled Strip At Masteel

Meeting the requirement of the cold rolled strip edge quality,which is the premise for the completion of strip processing.Traditional detection methods no longer meet the increasingly stringent requirements of cold-rolled strip edge defect detection tasks.The edge defect detection of cold rolled strip based on image processing technology is have a great significance on socity.This paper focuses on the key technologies of edge defect detection of cold rolled strip,including the edge image advanced preprocessing,defect sample acquisition,defect image feature extraction with which training classifier,conducting an real-time detection of cold rolled strip.The work completed is as follows:(1)Read a large number of domestic with foreign literature and relevant background research,clarify the steps and requirements of edge defect detection of cold rolled strip,and form the overall implementation scheme of the system.(2)Complete the advance preprocessing of cold rolled strip.It focuses on the east square method and minimum fitting rectangle,image morphology and horizontal gray projection,trilinear linear stretching and histogram equalization,Gabor filtering and Otsu.These are applied to correct angle,extract shear and tear,normalized image and segment defect.(3)The improved wld algorithm is used to extract features which used to train the classifier.The original data set is expanded by rotating and adding noise,which taken as the sample.Using timing,precision and recall as standard,SVM as classifier,competition with different feature extraction algorithms are carried out to verify performance of improved wld algorithm.(4)Based on Visual Studio 2013 and QT software,a cold rolled strip edge detection experimental system with three functional modules is designed.This system realizes the functions of edge defect sample acquisition,classifier construction and video real-time detection.(5)The performance of the edge defect detection system is tested by video of cold rolled strip.Taking the missed detection rate,false detection rate and time as the standards into evaluating the performance on tracking and detectingedge video,and the effectiveness of the edge detection experimental system of cold rolled strip is verified.