A 3D Omnidirectional Sensor For Mobile Robot Applications

Abstract

Catadioptic sensors, the association of a camera and a mirror, are powerful systems in applications such as robotics, tracking and surveillance where a large field of view is required. This paper deals with the calibration of a stereo catadioptric system used to provide a 3D reconstruction of a scene. This stereo bench is made of two panoramic vision systems, one above the other. Firstly, we present a brief summary of the catadioptric sensor theory and some of the common calibration techniques. Next, a classical analytical calibration method is applied on our system. As the main drawback of this method is to assume that the sensor respects a mechanical constraint, called the Single Effective Viewpoint Constraint, we propose another approach based on a bilinear interpolation of the calibration data set. Finally, the last part of the paper presents the results obtained, which highlight the behavior of the interpolation-based calibration method.