Design of a multi-view photogrammetric system based on low cost cameras for the 3D forensic face recognition

Abstract

The manuscript introduces a multi-view photogrammetric system designed to generate detailed 3D facial models to improve forensic face recognition (FFR). Developed in collaboration with the Rome Carabinieri Scientific Investigation Department (RIS), this study represents an initial feasibility study for creating a 3D mug shot with sub-millimetre accuracy utilizing low-cost cameras. To evaluate the system's performance, we conducted three series of experiments. Firstly, virtual environment testing using a 3D digital model to evaluate system performance. Then, real-world setup test using three different camera types to capture the 3D printed 3D model. Finally, some tests in real environment testing with human subjects to assess the ability to handle human skin in real environment. For the first two test, a quantitative analysis of the reconstruction error was performed by measuring the distance between the generated point cloud and the 3D reference model. Using a system based on Raspberry Pi HQ cameras, we obtain an 80% having less then 1 mm of distance from the ground truth. This research demonstrates the potential of low-cost photogrammetry for generating accurate 3D facial models, which could significantly improve the effectiveness of forensic face recognition.