Abstract

3D Palmprint recognition has become a promising alternative tool for resolving problems compared to the robustness of 2D palmprint recognition. Regarding robustness, biometric systems that use 2D Palmprint suffer from being attacked by using a fake Palmprint identical. Given this, the current paper introduces a new 3D Palmprint recognition approach. Firstly, a set of preprocessing techniques has been applied on 3D depth image such as Tan and Triggs method which can effectively and efficiently eliminate the effect of the low-frequency component with keeping the local statistical properties of the processed image. Then, Gabor wavelets have been employed to extract features. After that, the extracted features have been used as an input in the collaborative representation based classification with regularized least squares (CRC_RLS) to classify the 3D Palmprint images. To evaluate its performance, the proposed algorithm has been applied on the PolyU 3D Palmprint database which contains 8.000 samples. The experimental results successfully and greatly improve the recognition results, especially when, we use Tan and Triggs method for preprocessing and Gabor for feature extraction with CRC_RLS for presentation and classification. We achieve a significant recognition rate of 100 % in lowest Runtime which reflects the robustness of the proposed recognition system.

Authors

Abdelouahab Attia¹, Abdelouahab Moussaoui², Youssef Chahir³, Mourad Chaa⁴
Mohamed El Bachir El Ibrahimi University of Bordj Bou Arreridj, Algeria¹, Ferhat Abbas University, Algeria², University of Caen, France³, Ouargla University, Algeria⁴

Keywords

Three-Dimensional Palmprint, Biometric, Gaussian Difference Filtering, Gradient Palms, Weberpalms, Gabor Features, Self-Quotient Image Algorithm