Direct approach to the self-calibration of omnidirectional cameras

Omnidirectional cameras are a popular choice of visual sensors in robotics because their large field of view is well suited for motion estimation and obstacle avoidance. However their practical use is often burdened by the calibration phase that can be time consuming and require an experienced user. The contribution of this work is a simplification of calibration phase by providing a direct approach to the calibration of this kind of sensor. This approach tracks a planar region in the scene to find the camera´s intrinsic parameters. The only requirement for the user is to select a plane in the first image of the sequence. The algorithm needs no prior knowledge about extrinsic, camera, lens or mirror parameters. In order to assess the performance of the proposed method, we perform experiments with synthetic and real data.