期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2018
卷号：96
期号：18
出版社：Journal of Theoretical and Applied
摘要：Detecting unusual events in crowded scenes has drawn considerable research interest lately. In this paper, an unsupervised method that relies on a spatio-temporal descriptor and a clustering technique is presented to tackle this problem. We employ space-time auto-correlation of gradients (STACOG) descriptor to extract spatio-temporal motion features from video sequence. Following that, the K-medoids clustering algorithm is used to partition the STACOG descriptors of training frames into a set of clusters. The frame abnormality is defined by distances between the center of the clusters and the motion feature extracted by STACOG. We have conducted experiments on various benchmark datasets and the results show that the proposed method obtains comparable results: 98.48% AUC for UMN, and 92.13% accuracy for PETS 2009, at the frame level. In addition, fast computation time of our method that satisfies the demand of real-time processing.
关键词：STACOG; K-medoids; 3D gradient; Abnormal event detection; Visual surveillance
