An Automatic Tracking System Using A Single PTZ Camera(INTERNATIONAL Workshop on Advanced Image Technology 2008)

概要

論文の詳細を見る
In this study, we developed an algorithm for achieving automatic, stable, accurate, and real-time detection and tracking of moving objects, such as pedestrians in a scene, using a single Pan/Tilt/Zoom (PTZ) camera for image acquisition. For detecting moving objects, we proposed a new algorithm that combines two basic techniques; first, it updates the background image at the timing of each frame and then obtains a subtraction image between the observed input and the updated background images to execute background subtraction. In addition, we proposed using the optical flow data detected on each pixel in the present frame image to define a new updated background image. Such optical flow data were only obtained within the regions extracted as candidates for moving objects. In this paper, the detection accuracy and the tracking performance by the proposed algorithm were investigated by experiments in pedestrian tracking tasks.
社団法人電子情報通信学会の論文
2007-12-31