Research Of3D Information Acquisition Technology By Binocular Stereo Vision

Binocular stereo vision system simulates the principle of the3D imaging biological eye. It photographs the same3D object in the world from different angles by a couple of cameras, and calculates the3D information according to the geometric relations between tow cameras. Binocular stereo vision algorithms and technology gradually become the study focus, it is great significance in the accurate tracking, three-dimensional measurement, virtual reality and so on, has the very important research value. The innovation is put forward a new camera calibration method while is better robustness and high precision, and deduced the factors influencing the precision of the binocular stereo vision system.The key technologies in binocular stereo vision which this paper research on include camera calibration, stereo calibration, stereo matching and depth information acquisition. The main research work can be summarized as follows:(1) Research on the principle of cameras mathematical model and calibration, then presents a new camera calibration method based on planar pattern which is not same as the solution with other. In the proposed method, the optical centre is acquiesced to be the image centre, other initial parameters are solved with linear method and optimized separately and precisely in the non-linear model taking into consideration lens distortions. Experimental results on simulated data and real data show that the proposed method have parallel astringency, but better robustness in the noise environment and smaller reverse projection error.(2) Research on the epipolar geometry principle of binocular stereo vision, and then rectifies the couple images to be the head-up binocular stereo structure. In addition, the factors that affect the precision of binocular stereo system are derived and analyzed.(3) Due to the current binocular stereo application mostly based on texture feature detection methods, which can only express the3d information of object to some of the local points, so in order to meet the performance be able to reconstruct the whole three-dimensional space structure, The paper research on global stereo matching algorithm with high difficulty and challenging, then focuses on the analysis of the belief propagation algorithm based on Markov Random Field and programs to calculate the depth map.(4) To build experimental platform of binocular stereo vision and program system software, so as to calibrate the couple cameras, rectify the two images and eventually generate depth map by stereo matching on two images. In order to verify the effectiveness of the binocular stereo vision system, the3D information acquisition experiment is designed, and it show that the system has good accuracy.