Research Of Moving Object Distance And Speed Measurement Based On Binocular Vision

Vision is the main way to percept and cognize the outside environment. Recently,with the development of the computer, people wish that they can give it the ability toobserve the outside information, so that the computer can accomplish some tasksinstead of people.And this will promote the development of computer vision. At present,many domestic and foreign researchers have dedicated to the study of machine vision,in order to make the computer to simulate the working mechanism of human eyes betterin terms of understanding the outside information.Moving object detection is an important field of computer vision system study.And the aim of this paper is to investigate binocular vision system for distance andspeed of the moving object. In this respect, it designed a new system for analyzing themoving object on the basis of the binocular vision system, and experimental researchand results analysis had been carried.The main parts of this paper includes several points as follows:Firstly, in the aspect of camera calibration, it focused on several commoncalibration algorithm currently, and the problem of imaging model and calibrationprocess was analyzed. Furthermore, the main emphasis of this part was Zheng youZhang calibration algorithm widely used currently.Secondly, in the aspect of motion detection, the popular method,such as thebackground difference method, inter frame difference and optical flow method werestudied in detail theoretically. By analyzing the exsperiment results of various methods,it concluded the applicable occasions of various algorithms. For the actual needs, thispaper proposed a method of extracting moving objects using histogram based on Gaussmixture modeling. Comparing with other algorithm, the extraction effectiveness of thealgorithm was analyzed.Thirdly, in the aspect of extraction and matching characteristics, it focuses onstudying the method of popular ORB feature point matching theoretically, meanwhilethe experimental results were analyzed. Besides this paper put forward a set of matchingalgorithm based on epipolar constraint in the block matching for requirements, and analyzed the matching efficiency of the algorithm.Forthly, in the aspect of speed and distance measurement, it studied the triangleprinciple and method of speed measurement in binocular vision system, and analyzedthe algorithm accuracy by comparing the actual results.The experiments of this paper were conducted based on the OpenCV+VS2010platform and the results were analyzed. Finally, the results showed that the accuracy ofthe information obtained basically achieved the goal.