Calibration of multi-head panoramic camera using 2D checkerboard under rotation photography

Multi-head panoramic camera (MPC) with ultra-wide field of view and high resolution is widely used, but its calibration depends on high-precision 3D control information. In this paper, a MPC calibration method is proposed using one common 2D checkerboard and rotatable platform. Proposed method is distinctly divided into two stages: firstly, side camera calibration using rotation sequence images. With elaborate consideration of setup of coordinate reference frame of checkerboard and rotatable platform, one strict rotating photography equation is first established for the side-camera of MPC. Based on it, initial external parameter values of the image sequence outputed from any side camera of MPC in rotatable platform are given, and further optimized by taking the corner of the chessboard image and a small number of tie points between the adjacent views of side camera as the observation values. Secondly, center camera calibration based on multi-view geometry. The MPC space coordinate system is first established by moving the origin of the rotatable platform coordinate system to the geometric centroid of center of all side cameras. Based on it, the solid multi-view geometry relationship between the center camera and the side camera are used as the constrain as well as a large number of tie points obtained by rotating photography are used as the observation values to optimize the parameters of the center camera through bundle adjustment. The experimental results show that the proposed method can achieve high-precision calibration of MPC, only using one 2D checkerboard and a rotatable platform. It has a good application value because of its high-precision, high-degree of automation and low requirements for environment (independence on 3D control field).