Research On Online Detection Method For Surface Quality Of Continuous Casting Slab Based On Image Features

After decades of development, continuous casting technology has made great improvement.however, due to the impact of speed caster、flux used by caster and smoothness of the mold surface and other factors.Cases about defects of slab surface are still difficult to avoid.Through research and analysis to the existing detection methods for surface quality of continuous casting slab,on the basis of summing up the pros and cons of various methods,this paper innovatively presents an online detection method for surface quality of continuous casting slab based on image features,and designed an online quality monitoring system for surface quality of continuous casting slab based on the method described above, We conduct an effective verification to the above methods by the use of the system in the steel mills.Traditional detection methods for surface quality of continuous casting slab based on image depend entirely on intelligent algorithm of image recognition,the speed of identify easily is limited by algorithm complexity and performance,and due to oxide, chatter marks, scratches, etc on the slab surface,error recognition rate is relatively high.Based on the above factors,on this basis, this paper innovatively proposes a image detection method to conduct the slab surface quality monitoring,which mainly use artificial recognition method and intelligent identification supplemented to collection of images,and on this basis, the overall architecture of using this method and online monitoring system is proposed,this paper describes the hardware architecture and software architecture of the system in detail.The effect of slab continuous casting process to surface defects in continuous casting is analyzed.Meanwhile defects on the slab surface are classified and described.This paper sums up the different characteristics of each defect and factors between each defect and lays the foundation for the next intelligent identification of defects.According to the feature of defects on the slab surface,and binding the requirements of detecting system.This paper decides to use the Sobel operator of classic edge detection operators to identify suspicious defect region,extracts the suspected defect region image by the search algorithm of edge characterized point,and makes up the general high-dimensional feature vectors by extracting the Contourlet feature vectors and GLCM feature vectors of suspected defect region.The general high-dimensional feature vector is as the final texture feature vector of each suspected defect region image and inputs that into the support vector machine(SVM) model constructed by sample pictures,then has a classification to suspected defects. This paper introduces identification and location to the suspected defect in detail,and extraction,and how to build the SVM classification model.Using C++and Matlab programming method to develop the monitoring software of the system,applying artificial and intelligent recognition method that the system used,this paper describes the implementation process about online monitoring system for surface quality of continuous casting slab based on image features in detail,also explains the principle of artificial recognition, the principle of slab border recognition, the principle of image recognition speed and the principle image mosaic that the system involved.We have tests by using the system in the field of steel continuous casting production line.the results show that on-line monitoring method of slab surface quality based on image features that this paper proposed has higher recognition accuracy rate to slab defect image compared to other common methods and and has obvious advantages.Therefore, online monitoring method for surface quality of continuous casting slab based on image features is feasible.