Double Active Tracking Based On Binocular Vision

The application of double-eyes vision techniques for active track is the hot point for studying in computer vision technique filed those days. The key studies of this paper are composed of five parts, which are camera calibration, video capture, image pre-processing, edging detection, feature extraction, stereo correspondence and object space orientation.In the fist place, this paper give the simple introduction for computer vision and the description of key techniques for it . Then aiming at the camera calibration of the vision system, it presents several common calibration methods. After that this paper gives some algorithms including image capture and digital processing, image pre-processing, edging detection and presents the comparing of these algorithms. This paper also presents detailed analysis for feature extraction and stereo correspondence. And according to the characteristic of the experiment condition of this double-eye vision system, it presents the algorithm process of the characteristic points correspondence. And then combining the correspondence result with the space geometry relation of double-eyes vision 3-d reconstruction, this paper constructs the time-space coordinate of the tracking object and being tracked object and get the speed and direction of aiming object.Based on the analysis of those algorithms, using VC++ to carry them and at the same time fulfilled the program of the whole double-eyes active tracking system's algorithm.