Research On Camera Pose Estimation In Underwater Imaging Environment

The accuracy of camera calibration and 3D reconstruction in complex environments remains to be studied.When the camera takes pictures in the underwater environment,the light will be refracted and the imaging model will be changed.Under the pinhole camera,the underwater refraction imaging model was established by analyzing the propagation path of light and the influence of refractive index of different media on the propagation direction of light.By decomposing and simplifying the model,it was found that both intrinsic and external parameters of the camera could be calibrated directly in the water.Firstly,the infinite homography matrix under the refraction model and the image of the absolute conic in space under the underwater refraction model are analyzed,which is called the refraction image of the absolute conic,and its expression can still be transformed into a symmetric matrix related to the internal parameters by simple transformation.Then it is proved that there is still a constraint relation between the intrinsic parameters matrix of the images of two infinite points in the orthogonal direction under the underwater refraction model and the images of the circular points.For any circle with known diameter,the image of the center of the circle can be figured out by analyzing the duality of the image of the circle and the images of the circular points as well as the relationship between the center of the circle,and then the images of the orthogonal vanishing points and the circular points can be figured out to complete the solution of the camera parameters.Then the checkerboard grid is used to solve the internal and external parameters through the homography matrix.Due to the influence of refraction,different homography matrices correspond to different points,but only the external parameters are affected.Therefore,the intrinsic parameters can still be obtained by solving the equations.In addition,this paper also proposes a method to estimate the principal point,which can be calibrated step by step to achieve the purpose of optimal solution.Finally,the effectiveness and feasibility of the proposed algorithm are verified by simulation and real experiment results.