본문 바로가기
  • Home

An Object Tracking System using CUDA Based Parallel CAMshift Algorithm

  • Journal of Knowledge Information Technology and Systems
  • Abbr : JKITS
  • 2013, 8(2), pp.17-24
  • Publisher : Korea Knowledge Information Technology Society
  • Research Area : Interdisciplinary Studies > Interdisciplinary Research
  • Published : April 30, 2013

조지훈 1 LEE SANG GU 2

1한남대학교 컴퓨터공학과
2한남대학교

Accredited

ABSTRACT

In this paper, we implement a real-time robust image object tracking system for PTZ(Pan-Tilt-Zoom) cameras. For imageobject tracking, we use the CAMshift algorithm based on the HSV color image distibution of detected moving objects. AsCAMshift algorithm requires much floating-poing computations, image flickering and data bottleneck are sometimes occurredaccording to the CPUs. In order to overcome the drawbacks, in the proposed system, we parallelize CAMshift algorithm byusing GPU(Graphic Processing Unit) in the CUDA enviorment. The experimental result shows that the speedup factor has12~14 times faster as compared as CPU method. This system can be applied to an effective and faster image surveillancesystem for continuous object tracking in a wider area.

Citation status

* References for papers published after 2023 are currently being built.