欢迎来到天天文库
浏览记录
ID:37816213
大小:694.76 KB
页数:8页
时间:2019-05-31
《基于Fuzzy-ART神经网络的红外弱小目标检测》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、万方数据第34卷第5期系统工程与电子技术V01.34No.52012年5月SystemsEngincc“ngandEkctronicsMay2012文童编号:I001—506X(2012)05—085707基于Fuzzy·ART神经网络的红外弱小目标检测陈炳文1,王文伟1,秦前清2(1.武汉大学电子信患学院,湖北武汉430079l2.武汉大学测绘遥感信息工程国家重点实验室.湖北武汉430079)摘耍:针对现有背景抑制算法未能有效抑制背景而导致目标检测率低的问题,提出了一种基于模糊自适应共振理论(fuzzyadapth陀re80nancetheory,Fuzzy-ART)神经网
2、络的弱小目标检洲算法。首先,采用Fu盟y.ART神经网络结合Robinson警戒环技术,建立自适应局部空间背景模型。并以此分析像素点的背景模糊隶属度来抑制背景杂波,然后依据目标与残留背景杂波的空间特征采用模板均差法来突显目标,并提出基于行列模糊聚类的自适应分割算法采提取候选目标}曩后结合目标的运动连续性进行多帧轨迹关联从而检测出真实目标。理论分析与实验结果表明.谊算法能随背景的局部情况来自适应调节空问背景模型,从而自适应抑制背景杂波、突显目标.能有效提高信囔比.检潮出弱小目标。关键词:模式识别I弱小目标检测;模糊自适应共振理论神经网络;Robinson警戒环;自适应分割中图分
3、类号:TP391文献标志码:ADOI:10.3969/j.issn.100l一506X.2012.05.0lInfrareddimtargetdetectionbasedonFuzzy。ARTneuralnetwOrkCHENBing—wenl,WANGWen—wei。,QINQian-qin92(I.&^叩Z吖日cc加以打J行,or啪tfo行。肌knmi埘"的。w“kn430079,c^f邶l2.S细押Jf(删L4幻陀lD删,orh,om口lio撑凸gf以e£一疗g抽S“rw撕一g·M4p户i行g口以R硎o£fS阴“玎g,肌^口竹№fw”岫,鼽妇疗43DD79,儡l船)A.
4、bsll’act:Inorderto∞lvetheproblemthatthecurrentapproachescannotsuppressthebackgmundclut—terseffectiveIyandresuItinap∞rdetecti∞performance,anovelinfrareddimtargetdetectionapproachbased0nfu杞yadaptivere∞眦ncetheory5、Robinsonguafdtobuildtheadaptivelocalspatialbackgroundmodels.Withthesernodel3.thebackgroundclutter5aresuppressedaccordingtothedegreeoffuzzymatchbetweenpixelsandmod-eIs.ThenadifferenceaIgorithmbasedontemplateaverageisutiIizedtohighlightthetargetsaccordingtothespatialfeaturesoftargetsandresidu6、albackgroundclutters.Theproposedadaptivesegmentationalgorithmbasedonfuzzyclu8terofrowsandcolumn5isnextusedtodetectthecandidatetargets.Finally。thetruetargetsarefurtherdetectedbythemulti·f腿metrajectoryrelatedalgorithmbasedontheconsistencyoftargetmotion.TheoreticalBnalysisandexperimentaIre3ult7、sshowthattheproposedapproachcanadjustthespatialbackgroundmodeIsadaptivelyaccordingtotheconditionoflocalbackground,andenminatethebackgroundclutter8andhighlightthetargetseffectively.Itiscapableofimpmvingthesignal·to-noiseratioanddetectingthetargetsef-fecti
5、Robinsonguafdtobuildtheadaptivelocalspatialbackgroundmodels.Withthesernodel3.thebackgroundclutter5aresuppressedaccordingtothedegreeoffuzzymatchbetweenpixelsandmod-eIs.ThenadifferenceaIgorithmbasedontemplateaverageisutiIizedtohighlightthetargetsaccordingtothespatialfeaturesoftargetsandresidu
6、albackgroundclutters.Theproposedadaptivesegmentationalgorithmbasedonfuzzyclu8terofrowsandcolumn5isnextusedtodetectthecandidatetargets.Finally。thetruetargetsarefurtherdetectedbythemulti·f腿metrajectoryrelatedalgorithmbasedontheconsistencyoftargetmotion.TheoreticalBnalysisandexperimentaIre3ult
7、sshowthattheproposedapproachcanadjustthespatialbackgroundmodeIsadaptivelyaccordingtotheconditionoflocalbackground,andenminatethebackgroundclutter8andhighlightthetargetseffectively.Itiscapableofimpmvingthesignal·to-noiseratioanddetectingthetargetsef-fecti
此文档下载收益归作者所有