Research And Implementation Of Matching Technique Of Feature Points On 3D Human Mesh Model

With the development of the informative society,vision always plays an important role in information acquirement area.Saving visual information by pictures is an important advancement in human society.Traditional two-dimensional image,the evolution converting binary image to color image made another great progress,but two-dimensional images still can't meet the demand of informative society.Compared with the two-dimensional image,the three-dimensional model is the more familiar with the real world objects,which contains more features and a large amount of data,the characteristics is the same with shape features of many real object.Compared with the traditional two-dimensional image,the models have more vital advantages on high resolution and fidelity.Sincerely,three-dimensional grid images have drawn people's attention,related technologies are needed to process the 3D models.Non-rigid mesh model is a representative example in three-dimensional model field,it can bear a certain degree of deformation and bending strength relative to rigid mode,and which also has a great rich of changed gestures.Under these diverse transformation conditions,how to find the corresponding relationship between the same three-dimensional model,whether they come from the same three-dimensional type,at this time,we need to take certain science and technology to deal with the work,accurately speaking,the characteristics of the three-dimensional model Extraction technology.In this paper,making use of HKS SI-HKS and WKS descriptors to extract the surface features of 3D models respectively.The construction of HKS comes from the Laplace-Beltrami eigenvector,which constructs the HKS descriptor using the Laplace-Beltrami operator's eigenvalue system.In the meantime,we calculate WKS,which is similar to the HKS descriptor and extracts the detail features on the surface of the 3D model.The extraction of non-rigid mesh model feature points can effectively reflect the characteristics of the shape points on the three-dimensional model surface,greatly describe the three-dimensional model surface and reflect the essential properties of the three-dimensional model.The characteristic information of feature points is the most effective expression for 3D models.In this paper,by calculating the feature similarity of the model's surface points of interest,we can find the points of interest with higher similarity and obtain their matching relationship of feature points;furthermore,we can use three kinds of descriptors to calculate the feature state distribution of the model shape points respectively,then to analyze the characteristics of each descriptor,select the best performance descriptor to complete the matching of the feature points.