进化多目标优化的稀疏重构方法-研究

《进化多目标优化的稀疏重构方法-研究》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、万方数据AbstractCompressedsensing(CS)isanemergingresearchtopic.CSgenerallyincludesthreeaspects:thesparserepresentationofthesignal,theobservationmatrixandreconstructionalgorithm.ForCS,thesparserepresentationofthesignalisthebasis,designingtheobservationmatrixisthekeypoint,andreconstructionalgorithmi

2、susedforreconstructingtheimage.Themainobjectiveofthispaperistodesignagoodcompressedsensingimagereconstructionalgorithm.Atpresent,therearemanyreconstructionalgorithmssuchasgreedyalgorithm,convexoptimizationmethod.Forthefisttime,weputthethinkingofmulti-objectiveoptimizationevolutionaryalgorithmi

3、ntothecompressedsensingimagereconstruction,proposeevolutionarymulti-objectiveoptimizationbasedsparsereconstructionframework,andprovideadetailedalgorithmframeworkanddescriptioncombiningwithimage’swaveletstructuralinformationatthesametime.Themaininnovationofthispaperisasfollows:(1)Weproposeanevo

4、lutionarymulti-objectiveoptimizationbasedsparsereconstructionmethodandprovideadetailedalgorithmframework.Inordertosolveactualproblems,wedesignmulti-objectiveevolutionarycoding,consistentmutationoperatorandlocation-informationbasediterativehardthresholdingalgorithm.Wealsoverifythefeasibilityand

5、superiorityofthealgorithminthewaveletdomainbyaseriesofcomparativeexperiments.(2)Weproposeanewblockingandscatteringobservingway.Inthewaveletdomain,someblocksmaynotbesparseusingthetraditionalblockdividingapproach.Thenewblockingandscatteringobservingwayweproposesolvesthisproblembyallocatingthespa

6、rsitypressureoftheblockswhicharenotsparsetootherblocks.Weproposeorthogonalmatchingpursuitalgorithmanditerativehardthresholdingalgorithmbasedonthenewblockingandscatteringobservingway.Theeffectivenessandsuperiorityofthenewobservingwayandtheproposedalgorithmsareverifiedbyexperiments.(3)Weproposea

7、non-convexcompressedsensingimagereconstructionalgorithmbasedontheblockingandscatteringobservingwayandevolutionarymulti-objectiveoptimization,whichcombinestheadvantagesofboth.Wealsoprovideadetailedalgorithmframeworkanddescription.Because