Automatic Target Detection by Optimal Morphological Filters

在线阅读 下载PDF 导出详情
摘要 Itiswidelyacceptedthatthedesignofmorphologicalfilters,whichareoptimalinsomesense,isadifficulttask.Inthispaperanovelmethodforoptimallearningofmorphologicalfilteringparameters(Genetictrainingalgorithmformorphologicalfilters,GTAMF)ispresented.GTAMFadoptsnewcrossoverandmutationoperatorscalledthecurvedcylindercrossoverandmaster-slavemutationtoachieveoptimalfilteringparametersinaglobalsearching.Experimentalresultsshowthatthismethodispractical,easytoextend,andmarkedlyimprovestheperformancesofmorphologicalfilters.Theoperationofamorphologicalfiltercanbedividedintotwobasicproblemsincludingmorphologicaloperationandstructuringelement(SE)selection.Therulesformorphologicaloperationsarepredefinedsothatthefilter'spropertiesdependmerelyontheselectionofSE.Bymeansofadaptiveoptimizationtraining,structuringelementspossesstheshapeandstructuralcharacteristicsofimagetargets,andgivespecificinformationtoSE.Morphologicalfiltersformedinthiswaybecomecertainlyintelligentandcanprovidegoodfilteringresultsandrobustadaptabilitytoimagetargetswithclutterbackground.
机构地区 不详
出版日期 2003年01月11日(中国期刊网平台首次上网日期,不代表论文的发表时间)
  • 相关文献