LDBM-PFR: Local Directional Binary Pattern Mixed Filtering for 2D and 3D Palmprint Fusion Recognition

LDBM-PFR: Local Directional Binary Pattern Mixed Filtering for 2D and 3D Palmprint Fusion Recognition

Single-dimensional palmprint images are susceptible to interference from noise, lighting, and various factors, leading to information loss and subpar recognition performance. To address the problem, a fusion recognition scheme of 2D and 3D palmprints based on Hybrid Filter Local Orientation Binary Pattern (HFLOBP) is proposed. Firstly, the HFLOBP algorithm is employed to extract 2D features and 3D Shape Index (SI) features respectively to improve the recognition effect. Secondly, Principal Component Analysis (PCA) is utilized to improve the recognition efficiency, to reduce the dimension of 2D features and 3D Surface Type (ST) features of palmprint, to improve the recognition efficiency. Finally, the multi-dimensional palmprint features are fused by Double Weight Threshold Algorithm (DWTA), and classified and recognized by Collaborative Representation (CR). Experimental results conducted on the 3D palmprint database of Hong Kong Polytechnic University reveal that, in comparison with alternative classification methods, the average recognition rate reaches 99.98%, with an average recognition time of 0.64 seconds. The proposed method not only demonstrates a commendable recognition effect but also meets the real-time requirements, underscoring its practical application value.