基于混合高斯模型的摄像机移动检测

DETECTING MOVEMENT OF VIDEO CAMERA USING GAUSSIAN MIXTURE MODELS

摄像机的故意干扰检测技术是数字视频监控系统中一个重要组成部分。根据摄像机移动前后拍摄背景不同的特点,对基于混合高斯分布模型的背景消除方法进行了改进。包括:背景模型匹配只使用亮度信息,将高斯分布模型按权值、方差排序,动态调整采样频度,等等。实验表明,该算法分割效果较佳,检测的实时性大大增强。

Detection of intentional interference by video camera is an important part in digital video surveillance system.Based on the characteristic that the shooting backgrounds are different before and after the camera moving,in this paper we improved the background subtraction method which uses Gaussian mixture models.These improvements include that the background model matching only uses luminance information,Gaussian mixture models are ranked according to weight and variance,and the sampling rate is dynamically adjusted,etc.Experiments show that the algorithm proposed in the paper has good segment quality,and the real-time performance is greatly improved as well.