Research On Binocular Vision-based On-road Obstacles Identification And Detection

With the rapid development of the promotion of city and cars,Traffic safety problems are steadily on the increase.On-road obstacles identification and detection becomes a research hotspot at home and abroad.Research on binocular vision-based on-road obstacles identification and detection,research contents the construction principle of binocular vision system based on the data;calibration of binocular vision system and correction;the algorithm of binocular vision system;implementation and test results of binocular vision system.(1)And the construction principle of binocular vision system based on the data.This study focuses on the two basic models of binocular stereo vision system,that is non parallel binocular stereo model and parallel binocular stereo model,and on this basis,analysis of the implementation steps and depth measuring principle of binocular vision system;Then real and practical problems in obstacle detection,we built the hardware platform and software platform system.(2)Calibration of binocular vision system and correction.The main research of the technology of camera calibration and correction.First introduces the basic theory of camera calibration;Secondly,the advantages and disadvantages of different camera calibration through research,comparison and evaluation of the existing calibration technology;Then the calibration method based on MATLAB toolbox.For different calibration distance and baseline distance of the experiments and precision analysis.(3)The algorithm of binocular vision system.First briefly introduces the necessity of algorithm design,and then from the detailed analysis of the digital image preprocessing,image segmentation,region of interest Kalman filtering theory and the motion state estimation and obstacle detection and tracking etc..(4)Implementation and test results of binocular vision system.On image preprocessing,image segmentation,obstacle detection and localization of software design and implementation are discussed.The part of the work is mainly completed by Matlab7.0.Among them,the image preprocessing part of the Gauss operator for image smoothing and denoising.