Research On Dimensional Measurement Method Of Continuous Casting Slab Based On Binocular Stereo Vision

In the modern steel and metallurgical industry,continuous casting billets that have been cut create billet burrs on their end faces,and the methods of removing these burrs have a tendency to be more intelligent.The plasma robot equipped with a vision system replaces the traditional burr removal method,which can ensure a better burr removal effect.Among them,the vision system is able to position the continuous cast billets in real time and assists the plasma robot with the burr removal work.Relying on the National Natural Science Foundation of China(51374007)and the Natural Science Research Project of Anhui University(KJ2020A0259),this paper studies the continuous casting slab size measurement method based on binocular stereo vision,and the positioning and measurement experiments are completed with the laboratory continuous casting slab model.The main work of this paper includes:(1)construction of the vision system's software and hardware platform,and the research on the principle of binocular vision measurement.An introduction of relevant coordinate systems and camera imaging models in vision systems is given.Furthermore,based on the Open CV open source vision library,the camera is calibrated with Zhengyou Zhang's calibration method to obtain the internal and external parameters of the camera.(2)Feature point extraction.The dimensional measurement of the continuous cast billet model relies on "corner point" information,but the traditional corner point detection algorithms of ORB and FAST suffer from missed detections and corner point aggregation.Therefore,an improved FAST algorithm based on contrast adaptive threshold is proposed.The experimental results show that the improved algorithm can correctly detect corner points,make the position distribution of corner points more uniform,and effectively reduce the mismatch rate of feature points.(3)Stereo matching.In order to quickly and accurately match the feature points,an improved ORB algorithm is proposed.Firstly,the grey scale values of the pixels in the domain of the feature points are weighted,then the grey scale prime centroid method is used to extract the main direction of the feature points,and finally the feature points are described by string descriptors instead of traditional binary descriptors.The experimental results show that the improved method has greatly improved the accuracy of feature point matching and measurement accuracy,which meets the requirements of measurement experiments.(4)Three-dimensional reconstruction of feature points and size measurement.Firstly,the pair-pole constraint and RANSAC algorithm are used to screen the correct matching points from the matching feature points and ensure the one-to-one correspondence of the left and right image corners.In order to improve the measurement accuracy,an interpolation method is used to accurately fit the spatial scale and position of the corner point to obtain the sub-pixel position of the corner point,and the sub-pixel coordinates are used to reconstruct the three-dimensional information of the corner point.Finally,the 3D dimensional measurement of the continuous casting billet model is completed.The experiments show that fitting the sub-pixel coordinates of the corner points can greatly improve the dimensional measurement accuracy of the continuous casting billet model.By comparing the experimental measurement results with the actual size of continuous casting billets,the measurement accuracy meets the experimental requirements,which verifies the feasibility of the algorithm in this paper.Both real-time and accuracy of the algorithm considered,it can meet the measurement requirements of industrial sites.