The Research Of Calibration Technique In The Video Level Detection System

The video level detection system is a system with camera to get level imageinformation, processing the image finally to strike the liquid level. With theadvantages of strong interference, easy maintenance and non-contact, the system isbeing in broad and depth studied. It uses the two-dimensional images getted bycamera to reconstruct three-dimensional geometric information of objects, thusensuring the authenticity of the obtained two-dimensional image is the premise ofaccurate measurement results. However, by the lens manufacturing precision, visualprocessing deviation, the image will appear different degrees of distortion, which willaffect the measurement accuracy of the video level detection systems; In the sametime, the angle and distance of the camera with the horizontal are not the same in theon-site installation of camera every times, how to calibrate the liquid level on-sitefastly and accurately become a key problem need to solve. Therefore, it’s important tocorrect the image distortion and calibrate the liquid height before the detection ofliquid level.In this paper, on the basis of analysing in-depth the status and progress of cameracalibration technical at home and abroad, a suitable video level detection systemcalibration technology plan is propsed, which correcting the distorted image andcalibrating the liquid level on-site respectively, its main tasks are as follows:(1) In the image distortion correction aspect,this paper describes the type andmodel of the image distortion, analyses and presents two typical distortion correctionalgorithms, including the Tsai calibration method, Zhang Zhengyou calibrationmethod, and propose a correction method based on the image plane. The experimentalresults show that this method can effectively correct the distorted image, improve theaccuracy of video detection.(2)In the liquid level calibration aspect, Introduces and analyses in detail twocommon used calibration algorithm:The calibration method based on cross ratioinvariance and calibration method based on neural networks, and propose acalibration method based on image level detection.The experimental results show thatthe method can calibrate iquid level fastly and accurately, besides,the calibratedaccuracy is high.Finally, we do some program and simulation to the proposed algorithm, and achieve the above algorithm real-time on the designed ARM11embedded system.Theproposed algorithms and hardware platform have been assembled into "Embeddedvideo level detection system” principle prototype, which have been used in severalusers, practice proved that the system has good stability and high measurementaccuracy.