Abnormal Event Detection Using Trajectory Features

Abstract

In this paper, an abnormal detector is proposed using trajectory features. An intelligent surveillance system could provide not only the recording function but also the detection of abnormal activities. Trajectory feature is an effective feature for detecting the abnormal activities. Since the monitoring spaces are much varied, pre-defined trajectories are not available in all cases. In this paper, the video data of normal activities were collected and segmented for training the detector. The trajectory features of moving objects were extracted and represented as a normalized feature vector. A fuzzy self-organized map based detector, an unsupervised detector, was built up to detect the abnormal activities in real time. Experimental results are given to show the effectiveness and efficiency of the proposed approach. Finally, some conclusions are made.