Research On Robot Binocular Stereo Vision

Stereovision, an active brunch of robotic technology, is one important attribute of intelligent robot. Binocular Stereovision can be defined as the process of tri-dimensional reconstruction through the abstraction, identification, matching and interpretation of twain images gathered from identical objective, mostly include: video frequency trapping, camera calibration, image preprocessing, feature extraction, stereo matching and tri-dimensional reconstruction as well. It is through further research of binocular Stereovision that we can bring important enlightenment to multi-eye vision system and present supporting technology for .visual problem of intelligent locomotive robot, industry assembling robot, domestic robot and community servicing robot.Appropriate hard & soft structure can be favor of soundness of function, simplification of structure, cost reducing. Based on vision system composed of dual CCD camera and personal computer, we do discuss camera parameter calibration, stereo matching, object location and depth map construction in this dissertation, and so on.In the chapter of camera calibration, different camera models have been reviewed based on co-ordination relation, then conventional and self-calibration are detailed discussed respectively. Through stereo matching we can find image positions of identical object point in twain images. Characteristic selecting, criterions determination and algorithms achieving are three important procedures. Object location is defined as coming up with three-dimensional coordinate expression through binocular vision model analysis using geometrical relationshipDepth map generalization is the course of visualizing depth information. Deferent depth map's expression and four interpolating methods are discussed. Thus a whole robot binocular vision system is constructed.