Research On Billet Recognition And Position Detection Algorithm For Reheating Furnace Based On Computer Vision

At present,the walking beam reheating furnace is widely used in the steel industry and it becomes a vital system equipment for billet production.Therefore,in the continuous and rapid automated billet manufacturing industry,the operating status of walking beam reheating furnace is normal or not,directly related to production efficiency of the reheating furnace even production efficiency of steel factories.Usually the operating status of billet is monitored through the real-time monitoring screen,however,it is limited and unrealistic to judge the operating status of the billet timely and accurately by the staff with eyes.So in this thesis,based on the operation process of walking beam reheating furnace and running videos of billets,research on billet recognition and position detection algorithm for reheating furnace on computer vision is proposed,to achieve real-time monitoring of billet position changes in order to help staff to monitor the operating status of billet better.The main work of this thesis is as follows:(1)An adaptive threshold image denoising method based on wavelet transform is proposed to denoise the billet image in a targeted manner.(2)The captured video image has undergone a perspective deformation and we almost cannot determine the situation of the distal billet,so the method of image geometric reconstruction based on perspective transformation is adopted to process the image.Because the image is fuzzy and the contrast between billet edge and furnace background is not large,the method of image enhancement based on morphology is adopted to process the image.In order to get the billet edge completely,the method of edge detection based on Canny is adopted to process the image.(3)A straight line extraction method based on RANSAC algorithm is proposed to extract the front edge of the billet.The pixel distance from the billet to the furnace threshold is calculated by using the straight line parameters and the threshold position parameters,further the actual distance is obtained according to the calibration relationship between the pixel distance and the actual distance.In this way,the actual position of the billet in the furnace can be reflected by its actual distance.The experimental results show that the adaptive threshold image denoising method based on wavelet transform can not only remove the noise in the billet image,but also keep the edge information of the billet in the image;The method of image geometric reconstruction based on perspective transformation eliminates the perspective deformation of the image;The method of image enhancement based on morphology lays a solid foundation for detecting the edge of billet completely;The straight line extraction method based on the RANSAC algorithm can extract the front edge of the billet accurately,and finally the actual position of the billet inside the furnace is obtained,and the real-time monitoring of the billet in the furnace is completed.