본문 바로가기
  • Home

Pedestrian Detection and Tracking for Intelligent Video Surveillance

  • Journal of Knowledge Information Technology and Systems
  • Abbr : JKITS
  • 2014, 9(6), pp.699-705
  • Publisher : Korea Knowledge Information Technology Society
  • Research Area : Interdisciplinary Studies > Interdisciplinary Research
  • Published : December 31, 2014

Kwangseok Lee 1 김영섭 2

1경남과학기술대학교
2(주)시그널웍스

Accredited

ABSTRACT

In the development of detection and trace system, the algorithms for it are generally classified into three methods. These are background image based method, temporal differencing based method, and probability based method. Of them all, difference based method is used the most because it is easy for implement and efficient, and it is divided into recursive one and non-recursive one. Gaussian complex model is suggested in typical recursive method, and eigen background in typical non-recursive one.This research propose the trace technology to the detected object after selecting pedestrian as the region of interest in the multi-object environment contained progressive input video of the video collection device of the digital camera or CCTV. After this method separate object from background using the excluded MAMF (Modified-AMF) to the processing of the positively curved region in the typical background technique, AMF (Approximated Median Filtering), and choose region of interest to the image contained various objects effectively as combining the characteristic image using magnitude of the region distribution contained spatial properties and statistical properties, and improved trace of the detected object effectively as applying CAMShift (Continuously Adaptive Meanshift) algorithm to the selected region of interest for the improvement of the calculation speed.

Citation status

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