Toward Development of a Face Recognition System for Watchlist Surveillance

Area of Science:

• Computer Science
• Artificial Intelligence
• Biometrics

Background:

• Face recognition is increasingly deployed in uncontrolled, real-world surveillance environments.
• Automated systems struggle to reject previously unseen faces, unlike human perception.
• Surveillance requires recognizing specific individuals while rejecting the general population.

Purpose of the Study:

• To develop a face recognition approach that mimics human ability to reject unknown faces.
• To enhance automated surveillance systems for accurate identification and rejection of individuals on a watchlist.
• To address the challenge of handling previously unseen faces in open-world recognition scenarios.

Main Methods:

• Proposed an approach based on identifying decision regions in face space for target individuals.
• Generated borderline images by projecting inside and outside these decision regions.
• Trained a dedicated classifier for each person on the watchlist.

Main Results:

• Extensive experiments demonstrated the effectiveness of the proposed approach.
• The algorithm successfully handled previously unseen faces in realistic environments.
• Live system experiments validated the approach's performance in practical settings.

Conclusions:

• The developed method effectively addresses the challenge of rejecting unknown faces in automated surveillance.
• The approach shows promise for improving the accuracy and reliability of real-world face recognition systems.
• Mimicking human perceptual abilities offers a viable path for advancing face recognition technology.