MmLwThV framework: A masked face periocular recognition system using thermo-visible fusion

Area of Science:

• Computer Science
• Biometrics
• Artificial Intelligence

Background:

• The COVID-19 pandemic necessitates face masks, rendering traditional face recognition systems ineffective.
• The periocular region (area around the eyes) is crucial for recognition but offers limited discriminative features.
• Existing algorithms struggle with limited training data, a common issue in real-world applications.

Purpose of the Study:

• To develop a robust and lightweight periocular recognition framework for masked individuals.
• To enhance recognition accuracy by integrating thermo-visible features and mitigating noise.
• To address the limitations of pre-COVID-19 algorithms in low-data scenarios.

Main Methods:

• Proposed a novel Masked Mobile Lightweight Thermo-visible Face Recognition (MmLwThV) framework.
• Utilized ensemble subspace network classifier for improved feature discrimination.
• Integrated visible and infrared (thermo-visible) imaging modalities.

Main Results:

• The MmLwThV framework demonstrated improved accuracy compared to single-modality recognition.
• Effectively mitigated noise present in thermo-visible features, enhancing reliability.
• Validated efficacy on the WHU-IIP and CVBL masked datasets.

Conclusions:

• The proposed lightweight framework offers a viable solution for periocular recognition in masked scenarios.
• The integration of thermo-visible features enhances robustness and accuracy.
• The system's lightweight nature allows for easy deployment on mobile devices.