Active Target Tracking Using PTZ Camera

Human beings, whose vision is active, are able to move the head or body to change the viewpoint to recognise the object. Active target tracking system, which simulates the human vision, controls the operation of camera actively to track target based on the analysis results and requirements.In this paper, we designed and developed an active tracking system based on PTZ IPcamera which can detect and track object in real time. Its hardware is AXIS 231D+ IPcamera, software includes two modules: detecting module and tracking module. Detecting module can detect faces and bodies based on a cascade of boosted and Adaboost classifiers, while tracking module tracks the objects and controls the camera based on the results to keep them in field of view (FOV). It can control the parameters of the camera actively to track the target, and it also can be considered as a platform for comparing and testing active target tracking algorithms.Based on analysing the typical tracking methods, such as template matching, mean shift, histogram ratio and variance ratio, we integrated them in the active tracking system. We showed the experimental results, and evaluated their performance on static camera and active camera respectively. Based on the analysis, we proposed an improved part-based model for tracking, the results proved that, compared with mean shift, histogram ratio and variance ratio, the algorithm improved the veracity and stability of the system obviously.