Reconnaissance d’Oreilles Basée sur l’Apprentissage Profond : Améliorer la Précision et la Performance dans des Scénarios Réels

Abstract

Automatic recognition of individuals from ear images is a rapidly growing research field, competitive with other biometrics such as facial recognition and fingerprints. The ear offers unique and stable characteristics over time, which can be captured with a traditional camera. Our research explores ear recognition using various feature extraction models and classification methods. The first architecture employs SVM and KNN algorithms with LLBP and ALBP descriptors, achieving accuracies ranging from 85% to 98.50%. The second architecture uses convolutional neural networks (CNN) on processed images, achieving a remarkable accuracy of 100%.