| Steel plate is one of the most important industrial products,which is widely used in automobiles,ships and other fields.Defects on the surface of steel plate directly affect the quality of steel plate,but also affect the service life of the plate.Therefore,surface defect detection of steel plate is an indispensable procedure in steel plate production.According to the requirement of a steel plant in China,a set of inspection system design scheme based on machine vision is proposed for the finished steel plate produced by the steel plant.The system consists of three subsystems,namely lighting system,image acquisition system and image processing system.The first two systems are mainly the choice of hardware,and the last system is mainly the design of algorithm.This paper chooses the illumination mode which combines surface LED light source with bright field and dark field.In view of the large size of steel plate,it proposes to use multiple CMOS cameras to collect images separately,and to correct distortion and image mosaic.For non-uniform illumination images,brightness equalization method is used to correct them;for image filtering and denoising,bilateral filtering with edge-preserving function is used;for defect image segmentation,Canny operator based on edge detection and morphological processing are combined.After the defect area is segmented,the minimum outer rectangle of the defect area is made,and the location of the defect is located by the minimum outer rectangle.Through the above operation,the defect detection is completed.Finally,feature extraction and classification of defect areas are carried out.The extracted features include geometric features,gray features,Hu invariant moment features and texture features based on gray level cooccurrence matrix.Support vector machine is used to classify defect images.The average correct rate of classification after training of existing samples is about 84%,which can meet the needs of factories. |
PDF Full Text Request