Research On Location And Measurement Of Continuous Casting Slab Based On Binocular Stereo Vision

In the metallurgical industry,technological changes are leading the development of industrial automation towards industrial intelligence.In the hot continuous rolling process,plasma robot is used instead of traditional hammer cutter and scraper to remove the burr of continuous casting slab,which can achieve better burr removal effect.However,when the position and orientation of continuous casting slab have changed,the traditional teaching and playback method difficult to meet the requirements of robot operation.Therefore,the position of the continuous casting slab needs to be determined in advance to guide the robot to perform burr cutting.This paper studies the method of continuous casting slab positioning based on binocular stereo vision,and the experiments of positioning and measuring are completed on the slab model.The main work of this paper is as follows:(1)Study the measurement and calibration principles of binocular stereo vision.Build a positioning and measurement system based on binocular stereo vision.The calibration of the binocular system is realized,and the calibration results meet the accuracy requirements of subsequent measurement and positioning.(2)Binocular stereo vision positioning method based on feature point matching.According to the characteristics of the measurement points of the continuous casting slab model,the FAST corner detection algorithm is selected and improved,and the local binary pattern(LBP)is integrated into the feature descriptor,which improves the detection rate of effective measurement points and the accuracy of matching feature points.The threedimensional reconstruction of the measuring points is carried out based on the triangulation principle and parallax principle,which realizes the positioning and measurement of the continuous casting slab model.(3)Binocular vision localization method based on edge fitting.The ideal corner points of the continuous casting slab model are obtained by fitting the edges,and the corresponding corner points are matched by the principle of polar geometry.Finally,the positioning and measurement of the continuous casting slab model are realized according to the principle of triangular ranging.Subpixel edge points were used to fit the edges,and clustering of subpixel edges was achieved by an improved K-means method.Experimental results show that the proposed method can effectively improve the positioning and measurement accuracy of the continuous casting slab model.(4)Binocular stereo vision location method based on point clouds registration.Firstly,the semi global matching algorithm(SGM)is improved to improve the matching accuracy of the algorithm,and a denser disparity map of the slab model is obtained.Then the template point cloud and target point cloud are generated based on the disparity map and 3D size of the continuous casting slab model.Finally,the registration of template point cloud and target point cloud is completed by using NDT algorithm and ICP algorithm,and the continuous casting slab model is positioned according to the transformation relationship between the two point clouds.Experimental results show that the positioning method based on point cloud registration is feasible.