Information Extraction System Based On Binocular Stereo Vision Depth Of Objects

The 3D information of the real world including height,width and depth. Among of them the depth information or depth perception play a key role in many computer vision applications.Binocular stereo vision technology belong to a kind of passive ranging. Binocular stereo vision technology is modeled on human use of binocular depth perception clues information approach, implementation of three-dimensional perception. That is, the use of two cameras shooting the same scene from different locations, and then recover the depth information from the disparity.In this paper, we lay a foundation of intelligent robot crawling objects,visual navigation and target tracking through studying the depth information extracting system based on binocular stereo vision.This paper first introduces the purpose and significance of studying the binocular vision system and the current researching at home and abroad. Then the paper focuses on the camera calibration,image pre-processing,stereo matching and extracting depth information of system.1. Camera Calibration section researches the traditional camera calibration method,taking into account the experimental environment and experimental equipment, paper adopted the high accuracy and less manual intervention semi-automatic chessboard plane camera calibration method for the precise calibration.2. The image pre-processing section analysis the common image pre-processing algorithm and select the optimal algorithm with the experimental environment.Then,builded a binocular vision experiment platform and designed the programe of the image smoothing,image sharpening and edge detection based on VC++ programming language.3. The stereo matching section due to the traditional fixed window are-based stereo matching algorithm have high matching error rate, sensitive for noise and matching speed low, a fast and robust are-based stereo matching algorithm based on disparity gradient is proposed. In this paper a new similarity function is employed to improve the robust of matching and disparity gradient binding is used to reduce matching search space and improve matching speed largely. At same time, according to the disparity gradient range judgement image disparity smoothing regional and edge regions and choose the large window and small window respectively to improve accuracy of match.4. Extracting depth information section the paper first introduce the structural model of binocular stereo vision, then we respectively derived extracting depth information principle of the use of ray intersection method under on any environment of the binocular stereo vision and the use of triangulation method under the parallel binocular stereo vision. we extracted the depth information form the shooting stereo image pairs.At last,the paper do the experiments and ananlysis the relative error of the system.The results show that the system extract the depth information of objectives from 1m on stability and higher accuracy.