ROMJIST Volume 21, No. 2, 2018, pp. 193-205
Silviu-Ioan BEJINARIU, Ramona LUCA Analysis of Abnormal Crowd Movements based on Features Tracking
ABSTRACT: Abnormal events in public places can be detected in real time by analyzing the video sequences from video posts, web or surveillance cameras. Sudden changes of speed and/or direction for all people moving in the scene may indicate the occurrence of an abnormal situation and in some conditions also the position in which the event occurred. The detection and tracking of the people in the scene require high computing resources which make this solution difficult to be applied for real time analysis especially in case of crowded scenes. In this paper is proposed an analyzer of scenes participants moving patterns. The Scale Invariant Features Transform (SIFT) is used for features extraction and the Lucas-Kanade (LK) algorithm for optical flow computation. The main advantage is that it requires low computing resources being suitable for real-time processing. The results obtained by applying the proposed method are compared to those obtained by other approaches presented in literature and also to the subjective evaluation of human observers. KEYWORDS: Features tracking, Lucas-Kanade algorithm, Scale Invariant Features TransformRead full text (pdf)