基于三帧差分和连通性检验的图像运动目标检测新方法  

Novel Method of Moving Object Detection Based on Three Frame Differencing and Connectivity Checking

作　　者：吕国亮[1] 赵曙光[2] 赵俊[3] 

机构地区：[1]渭南师范学院数学系,陕西渭南714000 [2]东华大学信息科学与技术学院,上海201620 [3]西安电子科技大学电子工程学院,陕西西安710071

出　　处：《液晶与显示》2007年第1期87-93,共7页Chinese Journal of Liquid Crystals and Displays

基　　金：国家自然科学基金(No.60672026);中国博士后科学基金(No.2005037783)

摘　　要：图像运动目标检测是视频监控、数字电视等重要系统中的核心技术。文章提出一种利用梯度信息、三帧差分以及连通性检验等后续处理措施和自动提取图像运动目标的新算法。分析和仿真结果表明,该算法具有较小的运算量和较高的处理速度(对于320×240×24bit图像,可达15ms/帧),并且能够适应噪声较大和整体亮度突变的情况。Deteetion of moving objects in an image sequence is vital for applications of image processing, e. g. , intelligent video surveillance systems and digital TVs. In this paper, a novel algorithm for automated detection of moving objects was proposed and discussed. It is characterized by a fusion of some improving strategies including exploitation of edge gradients, three-frame-differencing, and connectivity checking. The results of theoretical analyses and computer simulation presented show that the algorithm has some advantages over its competitors, e. g. , less computation cost and higher processing speed （e. g. , 15 ms per frame for a image of 320 × 240 × 24 bits）, being more robust with noisy and luminancevarying image sequence.

关 键 词：图像序列处理 边缘检测 三帧差分 运动目标检测 

分 类 号：TN27[电子电信—物理电子学] TP391[自动化与计算机技术—计算机应用技术]