A Real-Time Ranging Method Based on Parallel Binocular Vision

Abstract

To address the problems that the ranging operation of binocular vision is complicated and the feature-matching difficult, this paper presents a real-time ranging method based on parallel binocular vision. Firstly, it is based on the single parameter Division Model to determine the distortion function of camera systems suffering from radial lens distortion. Secondly, the parameters of the parallel binocular cameras are calibrated with the proposed simple three parameters model. In the paper we propose that matching points are the centers of the circular marker in the left and right cameras, which are obtained by Adaptive Threshold Segmentation and Connected Component Analysis (CCA). Experimental results show that the algorithm only takes about 30 milliseconds and the ranging error could be controlled less than 5cm under the certain conditions. The results of experimental measurement on real data are presented and discussed.