Research Of Interest Point Extraction Of Three-dimensional Model

In recent years,the application of three-dimensional technology has drawn wide attention.At the same time,it has also become a hot area of computer vision research.The application of three-dimensional models has become more and more widespread,including medical,video,games,VR(Virtual Reality)and AR(Augmented Reality)in all fields.In the development of three-dimensional models,it is very important to extract its interest points.The interest points in the three-dimensional models can be used to simplify the model triangles,select viewpoints and match the three-dimensional models.For the extraction of interest points,the research on three-dimensional models is far less than 2D images so far,but different 3D models of interest points extraction algorithms have different application scenarios.SUSAN algorithm is a commonly used algorithm in 2D feature extraction.In this paper,by analyzing the geometric features of three-dimensional models and the principle of three-dimensional points extraction algorithm of interest,3D-SUSAN based interest points extraction algorithms which extend the 2D SUSAN algorithm into three-dimensional space is proposed.In order to evaluate the performance of the algorithm,this paper uses the eye tracker to collect the eye movement data of the observer on the three-dimensional models and build the algorithm based on the Ground Truth database to establish the eye point of interest based 3D model database.The main contents of this paper include the following parts:(1)Introduce the principle of the interest point extraction algorithm of other three-dimensional model and its realization process.(2)A 3D-SUSAN based interest points extraction algorithms is proposed.Firstly,the neighborhood of the vertex is selected,the curvature of the vertex of the model is calculated,and the vertices of the model are processed by Gaussian function.And then use 3D-SUSAN based interest points extraction algorithms to extract interest points.Finally,the interest points extracted from the three-dimensional model are de-dummy and non-maximally suppressed,and the performance of the algorithm and other algorithms is evaluated by using Helin's laboratory database.(3)In order to evaluate the point-of-interest extraction algorithm of 3D model and verify the validity of 3D point extraction algorithm,this paper establishes and realizes the interest points database based on the eye movement of the three-dimensional model.First of all,Stanford University and SHREC2007 three-dimensional model database into a multi-angle display of two-dimensional image sequence for the use of eye tracker playback,and acquisition of volunteers in the extraction of two-dimensional image sequence to watch the eye movement fixation point,and then the two-dimensional gaze point is mapped to the three-dimensional model to obtain the three-dimensional coordinate data,and finally the three-dimensional coordinate data is generated using the Ground Truth algorithm to generate the point of interest database.(4)Using point-of-interest 3D database based on eye movement to evaluate the point-of-interest extraction algorithm based on 3D-SUSAN and the other three-dimensional model interest points extraction algorithm,and further explore the difference with other databases.The experimental results show that the algorithm of point-of-interest extraction based on 3D-SUSAN is more advantageous than other algorithms in terms of accuracy.This paper also analyzes the point-of-interest database based on eye movement and the existing database,and concludes that the existing database is suitable for the field of model matching.The database built in this paper is more suitable for 3D model reconstruction.