Feature Extraction And Matching Of Three - Dimensional Ear - Point Cloud Shape

With the gradual improvement of Internet technology, network and people’s work, study and life are closely linked, network security issues become the focus problem that people pay more and more attention on it. Biometric recognition technology can avoid the defect of traditional identification methods which is easy to forget, to loss, and to damage. Biometric recognition technology improve the reliability of the identification mode. In recently, Human ear recognition become a new favorite in biometric recognition technology. Comparing with other biometric technology it has obvious advantages.In this paper, the 3D point cloud model for ear recognition technology has carried on the related research, the main work is as follows:(1) Introduce the classical ICP algorithm, and Sparse ICP, ICNP, EM-ICP three kinds of improved algorithm the basic theory and algorithm process, analyzes and compares the advantages and disadvantages of each algorithm. These four kinds of iterative closest point algorithm is applied to the 3D ear model registration, comparison of time, error and results.(2) Firstly, normalize the 3D ear point cloud model, adjusting the position of all ear point cloud models in the database, then based on the Iannarelli system to extract four local feature region of 3D ear model, matching with Sparse ICP algorithm on the local feature region of 3D ear point cloud model, according to the distance between the corresponding points to judgment the difference between ear models.(3) Using D2 shape distribution algorithm to descript the ear global shape geometrical features: Firstly, the 3D point cloud model is divided into several triangular, random chose a triangle, calculate a new random points on the triangle, then form the D2 shape distribution histogram based on the model. Using the least squares method to fitting a shape distribution curves for the shape distribution histogram. Based on the Minkowski L1 norm method of probability density function to compare different ear models. Through the experimental analysis results, D2 shape distribution algorithm cloud descript the shape of 3D point cloud model without the effect of resolution and position, it is suitable for 3D ear recognition.