Photogrammetry in Forensic Science: Accuracy, Reliability, and Scientific Integrity

Rooted in geometry and mathematics refined over more than a century, photogrammetry is trusted in high-precision industries like aerospace and film production. In forensic contexts, trained experts apply it using strict protocols, accounting for the limitations of source imagery, such as low-resolution CCTV or mobile phone footage, and report measurements with clearly defined margins of error. Measurements are anchored to known reference points within the image, such as verified objects or scale markers introduced during re-enactments. Courts in the UK and internationally have accepted photogrammetric evidence, with expert testimony regularly withstanding scrutiny due to its testability, peer-reviewed methodology, and adherence to recognised standards.

When an expert reports a measurement with an uncertainty, such as "±1.5%", this reflects a scientifically calculated margin, not a flaw. All measurements carry inherent variability, and photogrammetry is no exception. Sources of uncertainty include camera calibration, image quality, reference point accuracy, perspective distortion, and human variability. These are quantified through validation studies and blind proficiency testing, often benchmarked against known values obtained via laser scanning. The ±1.5% figure typically represents an expanded uncertainty at a 95% confidence level, meaning the true value would fall within that range 95% of the time. This approach aligns with international metrology standards. The margin of error must be fit for purpose. For example, a ±1.5% margin may be acceptable when estimating a person's height from CCTV footage, but insufficient for tasks requiring sub-millimetre precision, such as reconstructing a mechanism of injury.

To derive this margin, experts conduct controlled studies using objects of known dimensions under conditions that mimic real-world footage. They also account for real world variables, known as the "uncertainty budget", including lighting, lens distortion, and pixel selection. These are combined statistically to produce a single, scientifically robust uncertainty value. For instance, a height measured as 1.70m with a ±1.5% margin means the expert is 95% confident the true height lies between 1.6745m and 1.7255m.

Reverse projection photogrammetry is particularly useful in forensic investigations where suspect height estimation is required. This technique involves recreating the original camera setup and overlaying new footage or reference objects onto the scene. When the original camera or its specifications are available, this method allows for highly accurate height estimation, aiding in the inclusion or exclusion of individuals from a scene.

Reverse Projection – How Accurate Is It?

Studies comparing reverse projection with other methods such as PhotoModeler show that both can achieve height estimation errors under 2 cm, depending on the quality of the footage and calibration. However, reverse projection tends to have slightly higher variability in error margins, especially when camera resolution is low or the subject is far from the camera.

Limitations

Accuracy can drop when resolution falls below 960H (960 horizontal pixels) or when subjects are more than 36 metres away. Cluttered or poorly lit scenes can introduce uncertainty, and reliable results depend heavily on accurate replication of the original camera setup. These limitations are not flaws, they are simply factors that must be managed with care and expertise.

By clearly stating the margin of error and adhering to validated methods, forensic photogrammetry provides courts with reliable, transparent, and scientifically grounded evidence.

Cited Sources

1: OSAC 2021-S-0037 Standard Guide for Forensic Photogrammetry.
2: SWGDE Best Practices for the Forensic Use of Photogrammetry (2023).
3: UKAS M3003: The Expression of Uncertainty and Confidence in Measurement (2019).
4: Liscio, E., Guryna, H., Lea, Q., & Olverb, A. (2021). A comparison of reverse projection and PhotoModeler for suspect height analysis. PhotoModeler.