Toward Development of a Face Recognition System for Watchlist Surveillance.

Abstract in brief

• The paper discusses developing a face recognition system for surveillance that can effectively identify individuals on a watchlist in real-world environments with many cameras, while correctly rejecting strangers.
• The challenge is recognizing and tracking watchlist members among numerous other people, especially since the system needs to reject faces it has never seen before, similar to how humans do.
• The proposed solution mimics human perceptual ability by focusing on decision regions for each watchlist person and using classifiers trained with borderline images to improve recognition accuracy, with successful results in both tests and real-life environments.]