Facial recognition for disaster victim identification

Area of Science:

• Forensic Science
• Computer Vision
• Biometrics

Background:

• Disaster victim identification (DVI) traditionally relies on DNA, fingerprints, and dental records.
• Ante-mortem data for these primary identifiers may be unavailable or difficult to obtain after mass casualty events.
• Advancements in automated facial recognition and accessibility of ante-mortem facial images present a potential solution.

Purpose of the Study:

• To explore the feasibility of using automated facial recognition technology to support DVI.
• To evaluate the performance of a commercial facial recognition algorithm on post-mortem versus ante-mortem images.
• To assess the impact of image quality and cause of death on facial recognition accuracy in a DVI context.

Main Methods:

• Evaluation of a commercial-off-the-shelf facial recognition algorithm.
• Comparison of post-mortem images against a database of ante-mortem images (e.g., from identity documents).
• Analysis of performance across various operational scenarios, image quality levels, and causes of death.

Main Results:

• Facial recognition technology demonstrated value in supporting DVI processes.
• Performance varied significantly based on the quality of the facial images.
• Cause of death was also found to influence the accuracy of facial recognition matches.
• This represents the largest evaluation of facial recognition for post-mortem and ante-mortem images to date.

Conclusions:

• Automated facial recognition is a valuable tool for expediting disaster victim identification.
• Image quality and cause of death are critical factors affecting system performance.
• Further research is recommended to optimize facial recognition application in DVI.