JIITA

An Enhanced Loitering Detection Scheme with Looking-Around Feature

JIITA, vol. 1, no. 3, pp.44-49, Oct, 2017

DOI: Soon

An Enhanced Loitering Detection Scheme with Looking-Around Feature

Yonghwa Kim, Yoo-Sung Kim
Department of Information and Communication Engineering Inha University, Incheon 402-751, Korea

Abstract: An enhanced loitering detection scheme is proposed that uses a look-ing-around feature along with the moving pattern features which can be extracted from partial trajectories, to differentiate loitering of a per-son from normal walking in surveillance videos. The looking-around feature is extracted based on the change amount of the skin color pix-els in the person’s head area within the continuous video frames. With the looking-around feature, the enhanced scheme can detect loitering behaviors more accurately with 92% accuracy than the previous loiter-ing detection scheme of 88%, especially for loitering of a person who is looking around frequently to check surrounding circumstances as the person moves forwards.

Keyword:  looking-around feature; skin color ratio; loitering detection; classifica-tion; partial trajectory

Full Paper: Full Paper: