Research On The Problem Of Multi Extremum Intrinsic Symmetric Shape Matching

With the rapid development of artificial intelligence and its application in people's daily life,computer graphics has become one of the hot issues.As the basis of computer graphics,image matching plays an important role in face recognition,automatic driving and medical image.Shape matching is mainly divided into rigid matching and non-rigid matching.At present,the solution of rigid shape matching is mature,but there are still many problems to be solved in non-rigid shape matching.Among them,the problem of intrinsic symmetrical shape matching is one of the difficulties,because in the algorithm based on feature matching,the existing feature descriptors can not distinguish the feature points of the symmetrical part of the shape,which leads to the failure of shape matching,and the matching of multi extremum intrinsic symmetry shapes(such as ants)is more complex than the ordinary intrinsic symmetry shapes(such as human body),and the detection of symmetry points is more difficult.Based on the analysis of the advantages and disadvantages of the previous classical algorithms,this paper proposes a new algorithm for the matching problem of multi extremum intrinsic symmetry shapes.Firstly,the Heat Kernel sampling and the farthest point sampling are combined,and the fusion sampling is carried out to obtain the feature points with ideal number and position;secondly,the symmetrical point detection is carried out by two-step method.In the first step,the feature points are preliminarily classified by means of the Heat Kernel signature;in the second step,based on the asymmetric points obtained in the first step,the symmetrical points are classified again by calculating the difference of geodesic distance;Then,we use the one point matching algorithm based on the Heat Kernel signature to do the initial matching;Finally,the existing methods are used to adjust the symmetry flip.The numerical experiments on TOSCA and McGill database show that the algorithm improves the matching accuracy,stability and universality without considering the symmetry confusion.However,the experimental results show that the existing methods of adjusting the symmetry confusion are not good for the adjustment of multi extremum intrinsic symmetry shapes such as ants.This is also the future research direction of the author.