Research On Vision Detection Of The Hot Rolled Strip End’s Shape

The visual detection of hot rolled strip steel is mainly used in the Hot-rolled steel stripproduction line of the steel factory. When the Hot-rolled steel strip go through the roughroll, the head and tail of the strip will distort and extent for the uneven pressure of roughroll, resulting the temperature drops sharply and great damage to the finishing rolls. Thusthe irregularly shaped head and tail of the strip must be cut. In the production of the factory,the long shear line will definitely cause a waste of resources, and the short shear line willdefinitely affect the quality of the strip. Thus acquisition of the accurate shape of the stripand calculating the optimal cutting line are very important and valuable. This thesis aims atsolving the detection of the shape of the Hot-rolled steel strip. And four work are asfollowing:Firstly, in this paper, the binocular vision system is proposed. According to theproblem of shape detection in the hot-rolled steel strip production line of the steel factory, abinocular vision system including light source, camera lens, cameras, computer and otherunits is designed and a triangular shape algorithm of the visual detection is proposed.Secondly, the feature extraction method of the edge of Hot-rolled steel strip isproposed. For the extraction feature of the strip, this thesis designs edge extractionalgorithm based on image morphology, combining traditional image preprocessing, Theaccurate end shape of the strip can gained.Thirdly, using imaging principle of the camera and the characteristic of triangle, usingbinocular vision cameras an improved feature extraction method algorithm is designed,successfully obtaining accurate end shape of the strip. Using the Machine VisionToolbox(MVTB) of matlab, the modeling and simulation in the hot-rolled steel stripproduction line are realized. The simulation results are satisfying.Finally, some experiments are implemented in the hardware platform with the rigidstructure composed of the binocular vision cameras. The binocular vision cameras cangather and process images at the same time. The experimental results are satisfying, and thefeature extraction method and the triangular shape of the visual detection algorithm arefeasible.