Single Camera Full Range Of Stereo Vision System Optimization Design And Calibration

In the field of computer vision, such as 3D reconstruction and robot navigation, it is usually required to obtain the scene information as much as possible. The field of view of conventional camera is very limited, while images obtained by a single omni-directional camera are not enough for the depth computation of objects. The recent technology of omni-directional stereo vision, with the advantages of both omni-directional and stereo vision, provides a new means for an intelligent system to perceive its surroundings. Generally, two identical cameras are needed for a stereo imaging system. However, the size and weight of two cameras may exceed the limited payload of small robot. Therefore, an omni-directional stereo visual system with compact structure and light weight is necessary.A single camera catadioptric omni-stereo system is designed in this paper for the application in MAV visual system. The main contributions of this paper are as follows:(1) Structure design and optimization of the system. The system consists of one hyperbolic mirror, one hyperbolic-planar combined mirror, and one conventional camera; to satisfy the requirements in both size limitation and system performance, constraints are conducted and then optimization is performed for the computation of system parameters.(2) Calibration of the system. In order to use the images obtained by this system to recover the depth information exactly, calibration is required. In this paper, a calibration algorithm based on the equivalent projection model is proposed by taking sphere as the calibration target, and it is proved that at least two spheres are demanded to solve the camera parameters and the length of the baseline.(3) The simulation and verification for the designed system are completed by POV-Ray, and then the effectiveness of the calibration algorithm is also tested with the images created by POV-Ray.