Isometric 3D Shape Partial Matching Using GD-DNA

(整期优先)网络出版时间:2018-06-16
/ 1
Isometric3Dshapepartialmatchinghasattractedagreatamountofinterest,withaplethoraofapplicationsrangingfromshaperecognitiontotexturemapping.Inthispaper,weproposeanovelisometric3DshapepartialmatchingalgorithmusingthegeodesicdiskLaplacespectrum(GD-DNA).Ittransformsthepartialmatchingproblemintothegeodesicdiskmatchingproblem.Firstly,thelargestenclosedgeodesicdiskextractedfromthepartialshapeismatchedwithgeodesicdisksfromthefullshapebytheLaplacespectrumofthegeodesicdisk.Secondly,GeneralizedMulti-DimensionalScalingalgorithm(GMDS)andEuclideanembeddingareconductedtoestablishfinalpointcorrespondencesbetweenthepartialandthefullshapeusingthematchedgeodesicdiskpair.TheproposedGD-DNAisdiscriminativeformatchinggeodesicdisks,anditcanwellsolvetheanchorpointselectionprobleminchallengingpartialshapematchingtasks.ExperimentalresultsontheShapeRetrievalContest2016(SHREC'16)benchmarkvalidatetheproposedmethod,andcomparisonswithisometricpartialmatchingalgorithmsintheliteratureshowthatourmethodhasahigherprecision.