Abnormal Event Detection Based on Multi-scale Markov Random Field
In this paper, we present a novel unsupervised method for abnormal behavior detection, which considers both local and global contextual information. For the local contextual representation, we firstly divide video frames into local regions, then extract l
- PDF / 881,064 Bytes
- 11 Pages / 439.37 x 666.142 pts Page_size
- 103 Downloads / 206 Views
Key Lab of Intelligent Information Processing of Chinese Academy of Sciences (CAS), Institute of Computing Technology, CAS, Beijing 100190, China [email protected] 2 School of Computer and Control Engineering, University of Chinese Academy of Sciences, Beijing 100190, China
Abstract. In this paper, we present a novel unsupervised method for abnormal behavior detection, which considers both local and global contextual information. For the local contextual representation, we firstly divide video frames into local regions, then extract low-level feature such as histogram of orientated optical flow (HOF) and sequential feature which is composed of K temporal adjacent frames for each region. The global contextual feature encodes the statistical characteristics of those local features like orientation entropy and magnitude variance. An online clustering algorithm is introduced to generate dictionaries for the local and global features respectively. Then, for any new incoming feature, a maximum posterior estimation of the degree of normality is computed by multi-scale Markov Random Field (mMRF) based on the learned model. The proposed method is evaluated on hours of real world surveillance videos. Experimental results validate the effectiveness of the method, and the detection performance is promising. Keywords: Computer vision · Anomaly detection · Multi-scale markov random field
1
Introduction
Detecting abnormal behaviors in videos is one of the most promising fields in computer vision. It is receiving increasing attention due to its wide range of practical applications such as smart surveillance, suggesting frames of interest that should be analyzed by an expert, and summarizing the interesting content. However, there are still several problems in anomaly detection especially for the scene consisting of complex correlated activities performed by multiple people. Firstly, unusual activities seldom occur and the large intraclass diversity of unusual and usual activities makes them even harder to be predefined. The main paradigm for abnormality detection in videos recently is to extract features and to learn a model on normal samples from the video. So that anomaly is detected as the one fitting the model badly. Various methods may differ in the feature they used and the model they built. Secondly, the visual context for scene c Springer-Verlag Berlin Heidelberg 2015 H. Zha et al. (Eds.): CCCV 2015, Part I, CCIS 546, pp. 376–386, 2015. DOI: 10.1007/978-3-662-48558-3 38
Abnormal Event Detection Based on Multi-scale Markov Random Field
377
tends to change over time, which makes the incrementally updated process even more necessary. Based on these problems, several methods have been proposed. Specifically,[2], [4], [11] determine abnormality based on the trajectory for each object. However, trajectory is too dependent on the tracking algorithm and may be unreliable in crowd scenes. [1] proposes a simple approach that measures typical optical flow speed and direction for each local grid to determine anomaly. Yet this method di
Data Loading...
