Research On Biometric Recognition Method Based On Three-dimensional Contour Of Palm

This thesis introduces the process and method of biometric recognition based on the 3D contour feature of palm.It is based on the general process of biometric recognition,which includes four modules: biometric information acquisition,data preprocessing,feature extraction and feature matching.In the feature extraction and matching module,two feature recognition schemes are studied,which are recognition based on 3D Palmprint and recognition based on depth information.In the feature information acquisition module,firstly,the main technical schemes applied to 3D biometric measurement are compared and introduced,and the 3D measurement scheme of structured light based on multi frequency heterodyne and phaseshift is determined.According to the scheme,volunteers are invited to collect 300 3D data of 50 different palms.The data preprocessing module mainly realizes the data processing of point clouds,including triangulation,boundary detection,noise filtering and hole filling.At the same time,the palm 3D model is unified into the same coordinate system based on the feature point positioning.The palm 3D contour database is obtained by preprocessing the collected data.In the method based on 3D palmprint,the ROI of palmprint is obtained by feature points,and it is transformed into palmprint gray-scale image determined by depth information after equidistant sampling.Palmprint Feature extraction is realized by Gabor filter with optimized parameters.The result of filtering is integrated by competitive coding and the palmprint image is obtained as the data of matching.Feature matching is realized by calculating Hamming distance,and decision threshold is determined by DET curve.The recognition rate of the optimized 3D palmprint method in the test set is 98.2%.Because the 3D palmprint information is completely determined by the depth information,the 3D palmprint recognition method has better robustness than the traditional palmprint recognition scheme.In the method based on depth information,the depth information of palm is taken as the recognition feature directly,but the depth value matrix data obtained after ROI positioning is large,so this paper realizes the data dimension reduction feature extraction by principal component analysis.The main information of the original matrix is represented by the projection vector of the feature space.Based on the characteristics of extracted features and the number of data sets,this scheme adopts the recognition method based on support vector machine.Compared and optimized the parameters of SVM in model training,realized multi classification through one-to-one combination pattern,and on this basis,gave the flow of test recognition rate.The recognition rate of the method based on the depth information is 99.3%.Compared with the palmprint recognition scheme,the recognition rate is optimized.This scheme verifies the rich characteristics of the 3D palmprint feature information.