The study of an improved algorithm for binocular stereo vision based on intra-vehicular robot system

Abstract

Binocular stereo vision algorithm is an effective method to position for intra-vehicular robot system(IVRS) quickly and exactly, however, large number of image data and much time cost have brought difficulties to solve the real-time applications of IVRS for the general vision system presently. Therefore, a binocular stereo vision system is designed and developed in this paper for extensive application of space position. Through the establishment of mathematical model, system calibration and homonymy points matching, the high precision object's 3D coordinate for robot system can be acquired. An improved fast matching algorithm is presented for the system. In this algorithm, the aim matching region is greatly reduced through using epipolar constraint mutually between two images, which not only saves the calculate time, but also advances the matching precision, and then the high precision matching solution can be realized by the use of reference difference gradient rule. Finally, design feasibility of binocular stereo vision system is validated by IVRS, meanwhile, the vision system is optimized through error analysis. According to the experiment, it demonstrates the applicability and significance of the system for IVRS positioning.