Creation of Biometric System of Identification by Facial Image

Abstract

This article is dedicated to the creation of a biometric system of the identification by facial image. The conceptual model and the mathematical model based on image processing methods (Gabor and Daubechies wavelet transform) were developed, also steps of image processing in the program and the forming of the feature vector by calculating the image statistical characteristics were described during the analysis. Considering that the primary efficiency outcome of the system is the identification accuracy, which is performed by the system, it became necessary to carry out the experiments to determine these parameters, specifically the methods that are appropriate to use in the system. The results of the experimentation indicated that the system performs the most accurate identification outcome with the use of Daubechies wavelet transform for the image processing, the standard deviation and the variance for the feature vector forming, and Euclidean distance, the squared Euclidean distance or the Canberra distance as a metric of the image classification. Using these parameters, the created system performed the 92,5% accuracy of identification.