2D/3D image (facial) comparison using camera matching.

Area of Science:

• Forensic Science
• Anthropometry
• Computer Vision

Background:

• Facial comparison challenges arise from variations in image angles and distances.
• Anthropometric measurements are key for objective facial identification.
• Existing methods struggle with pose and orientation variations in forensic images.

Purpose of the Study:

• To evaluate a novel 3D-to-2D projection method for anthropometric facial comparison.
• To determine the feasibility of objective facial identification using standardized projections.
• To assess the reliability of comparing facial landmarks across different image acquisitions.

Main Methods:

• Utilized 3D laser scans and 2D images of individuals.
• Projected 3D facial data onto a 2D plane matching perpetrator pose, focal length, and aspect ratio.
• Performed anthropometric comparisons of seven fixed facial points for matching and non-matching pairs.

Main Results:

• The 3D-to-2D projection method could not reliably distinguish matching from non-matching facial pairs.
• Optimal image conditions (resolution, expression) did not improve differentiation.
• Non-matching pairs were indistinguishable from matching pairs under experimental conditions.

Conclusions:

• Current anthropometric approaches using 3D-to-2D projection show limited effectiveness for facial identification.
• The method's inability to differentiate under ideal conditions raises concerns about its forensic application.
• Further research with larger datasets and diverse facial comparisons is necessary to validate this technique.