Calibrating distance mapping of non‐SVP catadioptric camera of the soccer robot

Purpose – The purpose of the present paper is to propose a full model‐based method for distance‐mapping calibration for the non‐SVP (non‐single viewpoint) catadioptric camera of the soccer robot. The method should be easy to operate, efficient, accurate, and scalable to fit larger field sizes. Design/methodology/approach – The distance‐mapping model was first constructed based on the imaging principle. The authors then calibrated the internal parameters using the mirror boundary and used the mirror center to choose the correct pose from two possible solutions. The authors then proposed a three‐point method based on a unique solution case of the non‐SVP P3P (perspective‐three‐point) problem to solve the external parameters. Lastly, they built the distance mapping by back‐projection. Findings – The simulation experimental results have shown that the authors' method is very accurate even when there is severe misalignment between the mirror and the camera and that all calibration operations, except the calibration of a standard camera, can be completed in 1 min. The result of the comparison with the traditional calibration method shows that the authors' method is superior to the traditional method in terms of accuracy and efficiency. Originality/value – The proposed calibration method is scalable to larger fields because it only uses the boundary of the mirror and three feature points on the field, and does not need additional calibration objects. Additionally, an automatic calibration method that can be used during the game can be easily developed based on this method. Moreover, the proposed mirror‐pose‐selection method and a unique solution to the non‐SVP P3P problem are especially useful for a non‐SVP catadioptric camera.