欢迎来到天天文库
浏览记录
ID:36778729
大小:322.48 KB
页数:6页
时间:2019-05-15
《基于小目标孤立度的小目标检测算法》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、第35卷第12期光电工程Vo1.35,No,122008年12月Opto—ElectronicEngineeringDec,2008文章编号:1003—501X(2008)12—0013—05SmallTargetDetectionBasedonSmhllTargetIsolationDegreeLIULing-qiao,FUZhi-zhong,QIANWei,DENGZai-qiang,XIEWei(SchoolofComunicationandInformationEngineering,UniversityofElectronicScienceandTechnologyofChina
2、,Chengdu610054,China)Abstract:Theproblemofdetectingsmalltargetininfraredvideosequencewasaddressed.Insteadofbeingregardedasprominenceorhighfrequencypart,theinherentisolatedattributeofsmalltargetwastakenintoaccountandredefinedassmallisolatedregionsinimage.Basedonthisidea,anewmeas~ementcalledSmallTar
3、getIsolationDegree(STID)wasdevelopedforevaluatingthelikelihoodofapixelasthecenterofasmalltarget.BycalculatingSTIDineverypixelinimage,thebackgroundclutterwassuppressed.Thenanoveladaptivethresholdwasproposedtoextractthepreciselocationofsmalltarget.Incorporatedwiththetwomethods,asingleflamesmalltarge
4、tdetectionalgorithmwasbuilt.Itshighperformancewasthenprovedinaserialofexperiments.Keywords:STID;smalltarget;adaptivethreshold.CICNumber:TP751.1DocumentCode:A基于小目标孤立度的小目标检测算法刘凌峤,傅志中,钱为,邓再强,谢伟(电子科技大学通信与信息工程学院,成都610054)摘要:针对在红外图像序列中检测小目标的问题,本文提出了一种新型的单帧检测算法,该算法包括了两个部分:基于小目标孤立度的背景抑制和一种基于目标信杂比的自适应阈值分割。
5、其中,小目标孤立度为本文提出的一种衡量像素属于待检测小目标的量度,与传统方法中把小目标视作图像中高频或者局部突出的思想不同,该量度着重考察目标的“孤立”的特性。实验结果显示,该算法的背景抑制效果明显,尤其对运用传统算法容易发生抑制不充分的建筑物边缘的有很好的抑制能力,与自适应阈值算法结合后的单帧目标检测也取得了满意的检测效果。关键词:小目标孤立度;小目标;自适应阈值1IntroductionSmalltargetdetectingandtrackingisofgreatimportanceinmanyapplications.Variousmethodshavebeendevelopedi
6、ndecadesws1.Becausethesizeofthetargetisquitesmall(normallylessthan9pixelsx9pixels),thespatialprocessingtechniquesthatdirectlyutilizetheshapeortextureofthetargetitselfwilllosetheirvaliditywhenthebackgroundisoccupiedbycomplexclutter.Alternatively,somespecialatributesofsmalltargetshouldbeconsideredto
7、developadetectionmethod.Theattributeofsmalltargetcallbesummarizedintotwocategories:thespatialattributeandmotionattribute.Inpracticalsystem,thetwoaspectsareoftenusedtogethertoachieveahigherdetectingrate.Howeveqins
此文档下载收益归作者所有