A Single Color Camera Stereo Vision System

In this paper, a novel single color camera stereo vision system is proposed. Two planar mirrors are used to create double views (blue and red views). A dichroic filter (DF) is used to combine them and eliminate their interference. The double views can then be captured by a color camera through blue and red channels. When the DF transmits the red light, refraction would occur. During calibration, we separate the calibration process: calibrate the virtual red camera and the virtual blue camera in order, and then calibrate their pose relationship. The refraction is removed in this process. Moreover, when computing the 3-D position of a point, the measurement error caused by the refraction is also considered. In this experiment, the interference between the blue- and red-channels is shown to be negligible. We verified the proposed vision system on the standard spherical and cylindrical surfaces. It is shown that the measurement accuracy is improved when the effect of refraction is considered. In addition, the noise robustness of this proposed system is also tested. The measurement accuracy would not be affected severely, if the standard deviation of the uniformly distributed random noise is less than 0.035. Finally, the proposed system is employed to measure the profile of a flower model. The proposed system has potential industrial applications.

I. Introduction

Stereo vision system, as an advanced sensor, can be used for three-dimensional (3D) shape measurement [1], robotic navigation [2], [3] and structure motion capture [4]. Generally, a stereo vision system consists of at least two cameras and its working theory is based on the triangulation principle [5]. To begin with, calibration is necessary to obtain the intrinsic parameters of the cameras and the pose relationship between the cameras. To measure the 3D position of one point, the corresponding points between the images should be located. Subsequently, triangulation principle is employed to compute the 3D position according to the calibrated parameters.