A stereo vision system for an autonomous vehicle

Abstract

Several techniques for use in a stereo vision system are described. These include a stereo camera model solver, a high resolution stereo correlator for producing accurate matches with accuracy and confidence estimates, a search technique for using the correlator to produce a dense sampling of matched points for a pair of pictures, and a ground surface finder for distinguishing the ground from objects, in the resulting three-dimensional data. Possible ways of using these techniques in an autonomous vehicle designed to explore its environment are discussed. An example is given showing the detection of objects from a stereo pair of pictures.