Abnormal event detection in crowded scenes based on deep learning

详细信息    查看全文

• 作者：Zhijun Fang ; Fengchang Fei ; Yuming Fang ; Changhoon Lee…
• 关键词：Abnormal event detection ; Crowd analysis ; Saliency information ; Optical flow ; Deep learning
• 刊名：Multimedia Tools and Applications
• 出版年：2016
• 出版时间：November 2016
• 年：2016
• 卷：75
• 期：22
• 页码：14617-14639
• 全文大小：2,064 KB
• 刊物类别：Computer Science
• 刊物主题：Multimedia Information Systems
  Computer Communication Networks
  Data Structures, Cryptology and Information Theory
  Special Purpose and Application-Based Systems
  
• 出版者：Springer Netherlands
• ISSN：1573-7721
• 卷排序：75

文摘

In this paper, we propose to use the deep learning technique for abnormal event detection by extracting spatiotemporal features from video sequences. Human eyes are often attracted to abnormal events in video sequences, thus we firstly extract saliency information (SI) of video frames as the feature representation in the spatial domain. Optical flow (OF) is estimated as an important feature of video sequences in the temporal domain. To extract the accurate motion information, multi-scale histogram optical flow (MHOF) can be obtained through OF. We combine MHOF and SI into the spatiotemporal features of video frames. Finally a deep learning network, PCANet, is adopted to extract high-level features for abnormal event detection. Experimental results show that the proposed abnormal event detection method can obtain much better performance than the existing ones on the public video database.