Structure Suture Learning-Based Robust Multiview Palmprint Recognition.

Area of Science:

• Biometrics
• Computer Vision
• Pattern Recognition

Background:

• Low-quality palmprint images under unconstrained conditions degrade recognition performance.
• Traditional single-view methods struggle to capture strong palmprint characteristics.

Purpose of the Study:

• To propose a robust multiview palmprint recognition method (SSL_RMPR) for enhanced feature representation.
• To address challenges posed by low-quality and unconstrained palmprint data.

Main Methods:

• Introduced structure suture learning (SSL) to create an elastic nearest neighbor graph (ENNG) on reconstruction errors.
• Integrated a low-rank reconstruction term with projection matrix learning for improved robustness.
• Developed a method that reduces model complexity by avoiding extra structure capture terms.

Main Results:

• The proposed SSL_RMPR method comprehensively presents salient palmprint features from multiple views.
• Adaptive enhancement of discriminant palmprint representation was achieved by exploiting label information and latent consensus structure.
• The method demonstrated superior recognition performance on real-world palmprint databases.

Conclusions:

• SSL_RMPR significantly improves palmprint recognition accuracy, especially under challenging imaging conditions.
• The structure suture learning strategy effectively enhances feature discriminability and representation robustness.
• The proposed method offers a computationally efficient and high-performing solution for multiview palmprint recognition.