基于RGB颜色空间的减背景运动目标检测  

Moving Objects Detection Based on RGB Color Space and Background Subtraction

作　　者：徐以美[1] 郭宝龙[1] 陈龙[1] 

机构地区：[1]西安电子科技大学机电工程学院ICIE研究所,陕西西安710071

出　　处：《计算机仿真》2008年第3期214-217,共4页Computer Simulation

基　　金：国家863高技术研究发展计划项目(2006AA01Z127);国家自然科学基金(60572152);陕西省自然科学基金(2005F26)

摘　　要：在计算机视觉领域中,运动目标检测与分割是一个基础而又关键的问题。减背景法是其中一个比较经典和常用的方法,其难点在于如何获取背景以及实现背景的自适应更新。针对该问题,提出一种基于RGB颜色空间的运动目标检测算法,充分利用了图像序列在RGB空间中的变化特点,首先通过抽取帧图像进行背景重构,即对图像序列中每个像素点的RGB值进行排序后取中间值作为该点背景像素的RGB值;在此基础上引入学习率对背景进行自适应更新,然后在RGB空间中进行前景目标提取,最后利用数学形态学和连通性分析对结果进行后处理。实验结果表明,该算法快速有效、能够满足实时要求。Moving objects detection and segment is a fundamental and critical problem in computervision field. Background sub - traction is a classical and widely used method and its difficulties are background reconstruction and update. So an algorithm for moving objects detection based on RGB color space and background subtraction is proposed in this paper. The background is first reconstructed using an image sequence of N frames. The range of RGB values at each pixel in an image sequence is calculated, and the middle values are considered as the RGB values of background. The leaning rate is used for background update. Then the fore - ground objects are obtained in RGB color space. Finally, morphologic operation and connected component measurement are applied to the results to eliminate the wrong detection. Experimental results show that the proposed algorithm combines the advantages of veracity and runtime, and fit for real time detection.

关 键 词：背景重构 背景更新 运动目标检测 减背景 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]