Thermal face recognition under different conditions.

Area of Science:

• Biometrics
• Computer Vision
• Machine Learning

Background:

• Thermal face recognition is crucial for security and identification.
• Existing methods often struggle with varying environmental conditions and occlusions.
• Physiological thermal information offers a unique biometric modality.

Purpose of the Study:

• To propose a novel thermal face recognition method utilizing temperature information.
• To enhance face recognition accuracy and robustness under diverse conditions.
• To develop an efficient system using machine learning for thermal face identification.

Main Methods:

• Utilizing a Bayesian framework for face extraction from thermal images.
• Employing selected thermal points as feature vectors.
• Training a Random Forest (RF) classifier, an ensemble of decision trees, for recognition.

Main Results:

• The proposed method was tested under normal conditions, with added noise, and with occlusions like masks and glasses.
• Experimental results show the method's robustness against various challenges.
• Performance was compared favorably against convolutional neural network-based techniques.

Conclusions:

• The developed thermal face recognition technique is robust and effective.
• It significantly outperforms classic methods, requiring substantially fewer feature points.
• This method offers a promising alternative for biometric identification systems.