Calibration of Non-Central Conical Catadioptric Systems from Parallel Lines

Most camera calibration approaches require the use of calibration patterns like checkerboards. This also applies to Omnidirectional Catadioptric Systems. In this paper, we propose a new calibration method for Non-Central Conical Catadioptric Systems which only requires long and parallel lines which are plentiful in man-made environments. It is based on the extraction of lines from a single omnidirectional image, where we impose parallelism restrictions to obtain the analytical solution of the calibration parameters. However, the calibration of this kind of non-central catadioptric systems is noise sensitive. The proposal has been tested with simulations and then with synthetic and real images to evaluate the accuracy of the proposed method.