| Video tracking technique, a new research subject brought forth by the development and application of digital video, is widely used in many fields such as military and civil area. In recent years, studies in moving object detection and tracking in complex background and spatial position measurement and autotracking of dynamic objects have risen the attention of numerous researchers home and abroad. Thus they carry through a series of groping research.Aiming at the requirement of high-speed processing ability in video tracking part of high-speed TV system, this dissertation mainly studies in real-time object tracking algorithms. Above all, techniques related to video image pretreatment are discussed including image enhancement, image segmentation, object detection and so on. Then the principle of video tracking is introduced in detail. In the phase of object tracking, various tracking algorithms commonly used are analyzed emphasized on centroid and correlation algorithms. Applying the method of window-prediction and layered search, improved correlation tracking algorithm is proposed. Considering that only gray information is used in traditional tracking methods, the paper stresses on the tracking algorithms of color image. Using the color information, color image SSDA and tracking algorithm based on CamShift are put forward.This paper mainly carries through improvements in the aspect of tracking algorithm in order to choose fast and accurate target detection and tracking algorithms, thus to reduce operation, enhance real-time processing and improve the reliability and stability of video tracking system. Experiments have proved that the algorithms advanced in the paper are simple and available with good tolerance and applicability, and are of great value for the design of TV tracking system in the future. |
PDF Full Text Request