智能视频监控中入侵检测算法的设计与实现

为实现在高清视频监控中实时、准确的检测出入侵的人员或物体，本文设计一种新的入侵检测算法。 算法将混合高斯模型与改进的帧差法结合，并针对实际监控场景作出以下改进：当监控画面分辨率较高时，采用块均值建模替代原混合高斯模型中的单个像素建模提高算法检测的实时性；在监控场景发生光照突变时结合改进的帧差法避免因光照带来的误检测。实验结果表明，本文的算法处理速度快，准确性高，适用于实时高清的视频监控系统。

The design and implementation of intrusion detection algorithm in intelligent video surveillance

In order to achieve intrusion detection timely and accurately in the HD video surveillance, this paper designed a new intrusion detection algorithm. The moving objects detection of the algorithm based on improved Mixture Gaussian Model (GMM) and the improved frame differencing. Firstly, the algorithm use the blocking average model instead of the single pixel model in GMM, its can dramatically improve the detection rate. Secondly, by combining the improved frame differencing, the algorithm can avoid error detection due to illumination variation. Experimental results demonstrate that the algorithm is fast and can accurately detect the intrusion objects, it is suitable for the HD intelligent video surveillance system.