标题：Visual attention shift-based event alarming for intelligent surveillance
作者：Sun, Jiande ;Zhang, Likun ;Yuan, Hui ;Xie, Jiangchuan
作者机构：[Sun, Jiande ;Zhang, Likun ;Yuan, Hui ;Xie, Jiangchuan ] School of Information Science and Engineering, Shandong University, Jinan, 250100, China;[Sun 更多
来源：Journal of Computational Information Systems
关键词：Intelligent surveillance; Key frame extraction; Object tracking; Occurrence of event; Visual attention
摘要：As intelligent surveillance system has been widely used, the processing of surveillance videos is becoming more and more important. In this paper, a visual attention shift-based method is proposed to alarm the occurrence of likely-to-be-concerned (LTBC) events by key frame extraction. In the proposed method, the static and dynamic visual attention models are combined to extract the visual saliency ranges, according to the temporal changing of visual saliency ranges, the key frames are selected out to mark the occurrence of the LTBC events, and at the same time, the LTBC objects in the key frames are extracted and tracked in the former and latter frames. Experiments demonstrate that the proposed method can alarm the occurrence of the LTBC events with the extracted key frames correctly and the alarming is realtime, which is helpful to hold the dangerous event back or within limits in practice. The tracking results also show that the extracted LTBC objects are usually the main roles of the LTBC events, and they can be tracked in the video frames stably. Copyright © 2013 Binary Information Press.