Title: On the Enumeration of Frequent Patterns in Sequences

Year of Publication: Nov - 2014
Page Numbers: 40-44
Authors: Rajeb Akram, Ben Hamadou Abdelmajid and Loukil Zied
Conference Name: The International Conference on Artificial Intelligence and Pattern Recognition (AIPR2014)
- Malaysia


In accordance with the increase of the demands on wide-area surveillance to minimize the damage for incidents such as terrorism and riots, the importance of the sharing of video footage among local surveillance hubs is increasing to take action quickly and effectively. In this paper, we propose a system which can reduce the network usage between local surveillance hubs by changing bitrates depending on contents of the video feeds. The system was designed to support various camera configurations available in a real surveillance setup by using a middleware called ASCOT, on which surveillance systems can be developed by assembling different types of video analyses. As an evaluation, we developed crowd congestion detection, face detection and motion detection on top of the middleware and evaluated them with two footages taken in a real important premise, one from entrance and another one from corridor. Crowd congestion detection reduced 41% of network usage for the entrance video footage, and face detection reduced 46% of network usage for the corridor video footage where motion detection reduced 6% and 27% respectively. Furthermore, we confirmed that the system can apply these video analyses functions and flexibly combine these analyses for video streams.