| As a multi-disciplinary field, for many researchers, computer vision has been concerned a long time ago. New theories and methods are emerging with its wide range of applications in the industry. As an important part of computer vision,binocular stereo vision has important application in many fields.Currently, the airport surface surveillance technology has its shortcomings and limitations at home and abroad. As a low cost of new emerging technology, computer vision measurement has the potentiality of non-cooperative surveillance which can be used as a useful complement to airport surface surveillance. This project aims to build a binocular vision system to achieve the positioning of objects within a large field of view.Here, the following sections are completed mainly based on the existing research results. Firstly, according to a mathematical model, calibration expression about single parameter are given.Then calibration expressions about two parameters and three parameters are deduced theoretically based on the same model. Calibration expressions about single parameter, two parameters and three parameters are verified with the data collected in the experiment. Secondly, for the further verification of these parameters above, the reference coordinates calculated with the help of GPS and the coordinates calculated with parameters calibrated are compared. The result shows the rightness of the parameters within the range of allowable error. Finally, to calculate the position of targets in the images obtained by binocular vision system,feature matching work are completed. Epipolar geometry constraints improved the SIFT algorithm on image to increase the number of feature points matching while reducing the appearance of false matching points. So, Combined with epipolar geometry constraints, the algorithm is improved so that the number of correct feature points increase while the number of false matching points can be reduced based on the SIFT algorithm. |
PDF Full Text Request