联合熵流与边缘算子的红外运动目标检测  

Infrared moving target detection combined entropy flow and edge operator

作　　者：王忠华[1,2] 邓鹤[2] 刘建国[2] 陈忠[2] 

机构地区：[1]南昌航空大学信息工程学院,江西南昌330063 [2]华中科技大学图像识别与人工智能研究所,湖北武汉430074

出　　处：《华中科技大学学报（自然科学版）》2011年第1期11-13,23,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61071136);国家青年科学基金资助项目(40801164);南昌航空大学校立基金资助项目(Jy0651)

摘　　要：针对复杂背景下红外目标检测问题,提出一种联合熵流与边缘算子的红外运动目标检测算法.该算法分4个步骤检测目标:首先将熵流作为一种图像运动描绘子,通过熵流确定含目标运动的区域;其次运用Canny算子检测含背景和目标的边缘;然后联合两者捕获背景成分急剧降低、近似精确的目标边缘图像;最后计算边缘图像聚类中心,依据亮度、熵流约束进行区域生长,识别红外目标.对于海空背景下的红外目标图像,实验结果表明该方法能准确地检测、定位目标.Aimed at the detection of infrared moving target in complicated background,a novel detection algorithm combined entropy flow with edge operator was proposed.The algorithm was divided into 4 steps.Firstly,target motion region was determined through solving entropy flow which was regarded as a novel image motion descriptor.Secondly,Canny operator was used to detect clutter background and target edge map.Thirdly,combining two vision maps capture proximately precise target edge map,which was suppressed background components drastically.Fourthly,infrared target was finally recognized through calculating cluster center in edge map,and then using region growing.Experimental results indicate that the method can detect and locate target precisely in infrared image in sea-sky background.

关 键 词：运动目标 边缘检测 熵流 聚类 区域生长 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]