On Improved Single Viewpoint Constraint Calibration for Catadioptric Omnidirectional Vision

The single viewpoint constraint (SVC) is a principal optical characteristic for most catadioptric omni-directional vision (COV). Precise calibration for SVC is needed during system assembling. However, owning to the nonlinear distortion in the imaging system, the calibration precision based on ideal perspective imaging model is often poor. In this paper, a new calibration method of SVC for the COV is proposed. Firstly, an image correction algorithm is obtained by training a neural network (NN). Then, according to characteristics of space circular perspective projection, the corrected image of the mirror boundary is used to estimate whose position and attitude relative to the camera to guide calibration. Since the estimate is conducted based on actual image model rather than the simplified model, the estimate error is largely reduced, and the calibration accuracy is clearly improved. Experiments are conducted on simulated images and real images to show the accuracy and the effectivity of the proposed method.