一种改进码本的运动目标检测方法  

An improved codebook moving target detection method

作　　者：张振[1] 韩彦芳[1] 徐伯庆[1] 

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200000

出　　处：《信息技术》2013年第2期172-175,共4页Information Technology

摘　　要：针对运动目标检测中存在的阴影及背景噪声和树枝摇摆等问题,提出一种改进型码本模型检测方法。该方法对YUV空间分通道进行处理以解决阴影问题,针对干扰和噪声等影响,在背景差分图像的基础上,对视频帧进行二维网格区域划分,然后再对各个区域进行处理。与原始码本算法及经典检测方法的对比实验表明,该算法检测效果更好,并能满足实时检测需要。An improved codebook model detection algorithm is proposed to solve the problems of shadows, highlights and background interference like waving trees in object segmentation. With respect to shadow interference,this method deals with video sequences directly in the YUV color space, in view of background interference, a 2-D mesh-based region partition method is used and each patch is handled separately. The comparative results show that the method has a better detection effect than MOG and eodebook, and it achieves robust detection for different types of videos.

关 键 词：目标检测 YUV空间 二维网格 码本 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]