Moving Target Detection Tracking System Based On PTZ Active Camera

Active vision is to simulate the function of the eye. And people can make ajudgment about the external conditions, and then change the angle of view as theinterested object moves. When detecting the target, the traditional fixed camera can nottrack the target intelligentlly. At the same time, the range of its vision is limited.Therefore, if you want to accomplish a wide range of tracking, you may need morecameras. Without increasing the amount of cameras, PTZ (Pan/Tilt/Zoom, translation/rotation/zoom) cameras get the favor of people. Currently, active tracking technologyuses mainly the combined method of a fixed camera and PTZ camera. Fixed camera isused to detect and PTZ camera is used to track. In this paper, only using PTZ cameracan achieve detection and tracking.The main contents of this thesis are as follows:First of all, this paper proposed an improved target detection algorithm after weresearch common target detection algorithms, which is dual differential multipliedmethod. The experiments show that it can more complete extraction target contour.And then, this paper introduces common color spaces in image processing, whichare RGB and HSV color space. It tells the calculation of the color histogram as well asthe reverse projection. Then we study Kalman motion estimation algorithm, Mean Shifttracking algorithm and Camshift tracking algorithm. On this basis, this paper uses themethod of the direction and displacement prediction to solve the problem of missingtarget. And tracking algorithm is improved to some extent.Then, Face detection algorithm is studied. In order to achieve a good robustnessunder complex background, this paper adopts the new algorithm of AdaBoost facedetection with the risk which is putted forward by Ma Yong and some other people, andthat is CS-AdaBoost algorithm[1].And we apply the algorithm in the PTZ activetracking system as a post-processing. Firstly we use the moving target detection to filterout the non-target frame, and detect face in key frames with targets. We grade detection,and the algorithms is from simple to complex. Therefore it indirectly compresses videoframe rate, so that it increases the processing speed and achieves the real-time of the system.Finally, this paper introduces the PTZ active camera, analyses its characteristicsand the relevant agreements. We set up a moving target detection and tracking systemplatform based on the PTZ, describe the various modules of the system in detail. And anew method of target tracking is proposed. In this method, the movement of the camerais discretized to achieve the target tracking. When the camera is stationary, we conductthe target detection, and then according to the detected results, we can control the PTZcamera’s rotation and snap faces appearing in the process of tracking. The results confirmedthat the system had a good real-time detection effect.