Research On Calibration Technology Of Folding Reflection Camera

Catadioptric camera is widely used in 3D scene reconstruction,video surveillance,visual navigation and other fields because of its advantages of large field of view,simple structure and low cost.In this paper,the imaging model and error evaluation of catadioptric camera are studied.In this paper,the imaging structure and two existing imaging models of catadioptric camera are discussed,and the mathematical expressions of the two imaging models are derived and given.The first mock exam is based on the projection geometry theory.The imaging model of theodolite is presented between the catadioptric image points and the space straight lines,and the equivalence between the model and the spherical unified model is proved.In this paper,the hyperboloid mirror catadioptric camera is used for experimental verification,and Harris algorithm is used to extract the corner coordinates of seven chessboard images collected by the hyperboloid mirror catadioptric camera.Secondly,the first mock exam is carried out to calibrate the camera's internal and external parameters through the spherical unified model.In the theodolite model,the catadioptric camera can be equivalent to the traditional transmission camera,and the external parameters can be calculated by the traditional EPNP algorithm.Finally,the back projection errors calculated by the two models are compared.The first mock exam is that the average error of the back projection of theodolite model is reduced by 0.05 pixels than the spherical unified model,and the camera calibration accuracy is improved.Theodolite model the first mock exam of nonlinear spherical projection into equivalent form of linear projection,which can better use the existing position and pose estimation algorithm to get high-precision external parameters,and open up a new way of calculating the external parameters of catadioptric camera.