Research On Calibration Method Of Underwater Binocular Vision System Based On Refraction Model

The underwater computer vision has extensive application in Marine technology,such as: Submarine topography surveying and mapping,Marine resources exploration,Marine environment research,etc.As one of its branches,computer binocular vision consists of the following parts: image acquisition and processing,binocular camera calibration,stereo matching and three-dimensional reconstruction,etc.And as the basic link of computer binocular vision,the calibration of the camera plays an important role.The accuracy of camera calibration has great impacts on accuracy of stereo matching and 3D reconstruction.In this paper,for the special environment of underwater,combined with the existing underwater camera calibration method,some improvements are made to the underwater binocular camera calibration.Main works are as follows:Firstly,the imaging of underwater camera in special environment is studied.Because the imaging of camera underwater is different from single media imaging in air,light from the underwater object will go through “water-glass-air” three kinds of media and process multiple refractions,and then projects onto the camera imaging surface.However,there is no refraction problem in the air,and the corresponding model is no longer applicable,so it is necessary to reconstruct the imaging model for the refraction problem of underwater camera imaging.Secondly,as the calibration method in the air is no longer applicable,according to the problem of existing calibration method,based on the general refraction imaging model of underwater camera,by analyzing the geometrical relationship in the process of underwater object point project to the imaging plane through multiple refraction,the mathematical relation between the underwater objects point and the corresponding projection points on the imaging surface is established finally.Choose the absolute error(projection error)between the “projection imaging points” from the refraction model and the actual detection image points as the evaluation standard.Then solve the minimum value of the above established of error function in MATLAB,combining with particle swarm optimization(PSO)algorithm.Finally,the unknown parameters of the underwater binocular system can be obtained after several iterations.Finally,in order to verify the accuracy of the calibration algorithm,the software synthesis simulation and the actual underwater experiments are carried out.Using software to simulate the calibration board of multiple different attitudes underwater,and then the calibration method in this paper is used to calibrate the unknown parameters in the experiment,comparing the calibration parameter results with the ideal preset parameter values to test the accuracy of the calibration method.In the actual experiment,it is divided into two parts: The first is to calibrate the camera's intrinsic parameters in the air,using the Zhang Zhengyou calibration method;the second is calibration of refraction parameters for underwater cameras.Extracting calibration board corner coordinates and space point coordinates from underwater calibration board images,using the optimal calibration method established in this paper,and optimizing the refractive parameters of left and right cameras at the same time,finally realize the integrated calibration of underwater binocular refraction parameters.