Proceedings Article | 21 August 2013
KEYWORDS: Image processing, Panoramic photography, Video, Video surveillance, Computing systems, Detection and tracking algorithms, Image enhancement, Intelligence systems, Environmental monitoring, Logic
The Perimeter intrusion recognition technology has slowly become an indispensable function in the intelligent video
surveillance system. The existed always use the multiple video acquisition nodes to respectively control a monitoring
area and each node alarm independently. However, the existed solutions are difficult to avoid the existence of monitoring
blind area, and can’t suitable for the perimeter environment with irregular outline, and at the same time, because of the
too many nodes, it inevitably decreased the overall accuracy of intrusion recognition system and increased the cost of
system. To avoid the above defects, this paper mainly talks about the following three aspects. Firstly, we used the fisheye
lens as the optical system of video acquisition node, and it evidently enhances each node’s information acquisition ability.
And in this way, we just need to decorate a small amount of video acquisition node to get no blind area environmental
information of the perimeter when against a larger monitoring situation. Secondly, due to the inexistence of blind area,
the system will have enough video image information to generate the 360 degree panoramic image for monitoring
environment, and finally the system server collected the wide-angle image information to splice into the panoramic video
image. Finally, the system will use the panoramic image to complete the intrusion behavior recognition, thus we can
effectively avoid the parallel computation in many nodes independently invasion of recognition, and this can greatly
reduces the dependence for the multiple CPU operation platform and enhances the reliability of the system. The field test
results show that, with the help of this paper’s solution, the perimeter of the invasion of recognition system can
effectively avoids the recognition of blind area. In the same recognition algorithm and same level delay premise, it
greatly reduces the monitoring system server configuration requirements, especially for the requirements of the number
of processors. At the same time, the user also can directly set no blind area of monitoring perimeter and other parameters
in the panoramic image, so it enhances the flexibility of user operation and simplifies the design of the system user
application program greatly.
