A Stereovision System for 3-D Perception

Abstract

The results of investigation on a stereovision system for 3-D perception are presented. A 3-D space data can be reconstructed from a pair of 2-D images, the basis of stereo vision and computation equations are given. Also error of depth estimation and ways to minimise it are surveyed. The systems’ hardware design and the ways to synchronise cameras to avoid incorrect map reconstruction due to asynchronous frame capturing are presented. The results of experimental investigation: measured and extrapolated distances at certain disparities using 6 mm and 2.5 mm lenses, error of distance measurement distribution are presented. The results show that usage of longer focus lenses is more useful at long range perception. Also the results show that the system surveyed is suitable for autonomous robot navigation system. Ill. 7, bibl. 8 (in English; summaries in English, Russian and Lithuanian).