Research On Shape-aware Technology Of Quenched Steel Plate Based On Machine Vision

Steel plate is one of the most important steel products in China and an important part of the national economy.In recent years,with the continuous improvement of the quality requirements of steel plates,the development of steel plates has been driven towards automation and intelligent control.Quenching treatment can improve the comprehensive mechanical properties of the steel plate.The shape control of the quenched steel plate has always been an internationally recognized problem.Straightness is an important indicator of the quality of quenched steel plate,and on-line detection of flatness has been a research hotspot.So far,the straightness detection equipment of Chinese steel production enterprises basically relies on imports,and the independently developed straightness detection equipment is basically in the experimental stage.Therefore,the development of the straightness detection equipment can not only break the monopoly of foreign technology,but also have great significance for improving Chinese steel technology,and helping the steel plate production to achieve closed-loop control.The straightness detection equipment based on the laser triangulation method has the advantages of high precision,simple maintenance,and low price.In this paper,theoretical analysis and experimental research are carried out around the laser triangulation principle,image preprocessing method,edge detection method and straightness detection factors,as well as the construction of the detection platform and the design of the software system.The main research contents and results are as follows:(1)The image preprocessing method was studied and verified with the actual shooting results.Image preprocessing includes lens distortion removal,sharpness control and local adaptive ROI setting.Zhang Zhengyou calibration method is selected to remove lens distortion.According to the characteristics of laser stripe,a new definition evaluation function is proposed,and a variable step size definition control method is designed based on this function.The local adaptive ROI setting method is proposed based on the Otsu method and the concept of bright spot density.(2)Laser fringe edge detection is the core of laser triangulation.Improving Canny edge detection algorithm relies on the characteristics of laser stripe.The filtering method of laser fringe image is selected by PSNR value,the first-order differential operator is selected by the non-maximum value suppression results,proposing a threshold adaptive algorithm based on he characteristics of the edge of the laser stripe,the problem of different intensity of different regions of laser stripe is solved by local adaptive threshold,proposing an edge breakpoint reconnection method based on the edge feature of laser stripe.(3)Studing the best calculation method of height-pixel position curve,then exploring the effects of laser incident angle,camera shooting height and surface roughness of the measured object on the straightness detection accuracy and summarizing how to set various parameters in the scene.Finally,exploring the effects of the number of pictures,the number of corner points,grid size and lighting conditions on camera calibration accuracy.(4)Selected the equipment relies on the requirements of the project,and built the straightness detection test platform.The straightness detection software system was designed according to the proposed image preprocessing method,laser stripe edge extraction method and straightness detection method.Some steel plates were selected to simulate the operation results of the experimental platform and software system under strong light and a large amount of water vapor environment.