Ptz Camera-based Moving Target Active Tracking Technology

Computer vision refers to a kind of simulation to biological vision, by using computers and related equipments. Its main purpose is to gain three-dimensional information from a certain scene through operating collected pictures or videos. And active vision has been applied more frequently into computer vision, since it can overcome the inherent defects in computer vision by intelligently shifting the visual angle of active video. For example, due to the limitation of visual angle, it is difficult for tracking targets in a large scale or from multiple angles. But PTZ (Pan/Tilt/Zoom) camera is better choice, with not increasing number of cameras, can be a guarantee of tracking certain targets in a larger scale. With the wide application of PTZ camera, the study of active tracking technology based on it has become popular.The purpose of active tracking is to involve the moving target within a camera's vision with the video following target's movements all the time. Currently, there are two cameras used in most of the studies on active tracking technology, a fixed camera as the chief and a PTZ camera as the auxiliary. The chief camera is used to analysis the whole visual scene, operate pictures, calculate the position of a target and produce controlling instructions. And then the auxiliary camera follows these instructions to record video, fix targets in tracking and so on. It is rare to use PTZ camera on its own to actively tracking a target. Mostly methods, like using artificial nerve net or discretizing movements of camera, are mainly applied. But they are difficult to be operated in time owing to the large amount of calculation and the sophisticated operation.In this thesis, the author mainly introduce his/her study on the active tracking technology which is on basis of PTZ camera's moving targets, taking the video surveillance as background. Based on the problems found in solely PTZ active tracking, the author offers a new method which divides the process of active tracking as well as discretizes movements of cameras. According to this method, the process of active tracking is divided into two parts. At first, the moving target is tracked when the camera remains motionless and then the result of it is used to control the movement of camera. The process of camera movement is discretized at the second part, at the beginning of the location of the object in central region and the goal set out to control the camera center area, and at the rest of time the camera remains motionless. A high efficient and accurate method of calculation named Mean Shift is applied to track moving targets, since the process of tracking is essential and the control of video entirely depends on it. However, the position of moving targets will greatly change after the video starting working. It will move beyond the searching scale that Mean Shift can reach. So the changed position of target needs to be fixed by modified and then to be used to update Mean Shift searching system to go on tracking. Through experiments, this method has been proved to be both feasible and capable to well control the video in tracking targets timely and actively.