家庭安全监控的实时异常检测与分类算法

Real-time algorithm for the detection and classification of the exceptions in the home security surveillance

同时具有好的实时性和稳健性是基于视频的安全监控系统异常检测算法的难点问题 .在家庭安全监控系统中同时利用像素的亮度分量和色度分量检测变化像素 ,解决摄像镜头由于大尺寸对象闯入引起的图像背景变化造成运动对象检测失败的问题 ;提出一种实时运动对象分类算法 ,综合利用开关灯检测、运动距离分析和皮肤探测多种方法 ,区分是否出现影响家庭安全的事件 ;提出 3关键值背景维护算法 ,用于消除背景振动像素 ,减少虚假报警 .实验表明 。

Fast and robust detection of moving objects was difficult for a security surveillance system based PC. By combing the illumination and the chroma in the YUV color space, the problem was solved that a big breaking in object causes the background pixels to change its illumination. A fast method for analyzing and classifying the moving objects was introduced, based on some detection such as light switching, moving distance and skins. In order to eliminate swaying objects in the background and to reject false alarms, a new method of background maintenance was proposed.