Research On Feature Extraction Algorithm Of 3D Model And Annotation Method Of Cultural Relic Fragments

The rapid development and wide application of 3D technology provide rich technical means and digital resources for the fields of engineering design,cultural relics and archaeology etc.How to use these resources reasonably can effectively improve efficiency and reduce costs,the study of feature extraction method is the key.The research results have great theoretical significance and practical value.Based on a number of National Natural Science Foundation Projects,this paper analyzes the existing 3D feature extraction methods and focuses on the characteristics of 3D models with complex geometric features and complex structures.The methods of topological features,differential geometry and local salient features are comprehensively utilized.The key research is carried out on the methods of 3D model feature extraction,shape analysis,and it is applied to the classification and identification of computer-assisted broken cultural relics.The research contents and innovations are presented as follows:1.A 3D model feature extraction method based on geodesic distance and normal angle is proposed.According to the associated attribute information of the 3D model vertices,the geodesic distance reflecting the topological properties is combined with the normal vector reflecting the geometric variation.First,the sampling points are extracted from the surface of the model according to the size of the triangular patches.Secondly,the geodesic distance between the sampling points and the normal vector angle between the corresponding points are calculated.Constructing a feature matrix based on the probability of statistical angle and geodetic distance value.Finally,use the distance calculation formula to get the similarity.Experiments show that the extracted features can effectively reflect the topological difference of the model,and it is better than the Euclidean distance method in terms of recall rate and precision.2.A 3D model feature extraction method based on surface type distribution is proposed.First,calculate the curvature information of the 3D model surface,and then analyze the local surface geometry according to the curvature to determine the type of the surface.Secondly,combine the surface type and the distance distribution to obtain the probability of different surface types,and construct a distance and surface type distribution feature matrix.Finally,the similarity comparison is performed by comparing the feature matrices.Since the feature not only contains the shape information of the 3D model,but also contains the geometric properties of the local surface.Therefore,the method has good descriptive ability for models with inconspicuous topological features but complex surfaces.3.A local feature extraction method based on surface type for relic fragment model is proposed.Apply surface type information to the local feature extraction of the armour model.First,the local surface types are calculated.Secondly,the local geometric features of the model are obtained by using the region growing and expansion search methods according to their type features.Thirdly,the local feature is combined with the prior knowledge to create a template library.Finally,the classification of the armour model is achieved by comparing the similarity between the template and the local geometric features of the model to be detected.The experimental results show that the method has a good recognition effect on the local regularity of the surface of the armour model.4.Based on local salient features,a method for labeling parts of relic fragment models is proposed.In order to extract the "meaningful" local features on the fragment model of different parts,an improved local salient function is introduced.First,the curvature index of the model vertex is calculated.The median filter and the mean filter are used to process the values.Secondly,iteratively generates local feature patches using quadratic surface fitting technique.The salient value of the patch is calculated by introducing the salient function based on the characteristic patch curvature index extreme value and the patch area as parameters.Then the local salient shape features of the fragment model are obtained.Finally,the part-in-whole method based on the fast exclusion strategy compares the model in the sample library with the fragment model to be labeled,and realizes the part labeling of the fragment model according to the calculation result.Experiments show that the method improves the accuracy of classification and labeling of cultural relics model parts.