U.S. flag

An official website of the United States government, Department of Justice.

Optimization and Statistical Testing of Experimental and Computer Methodologies for Evaluation of the LIT-MIA Thermal Imaging Method for Recovery of Defaced Serial Numbers

NCJ Number
304649
Date Published
2022
Length
74 pages
Annotation

This is the Final Report on a research project that sought to optimize and statistically test experimental and computer methodologies for evaluation of the lock-in thermography (LIT) coupled with multivariate image analysis (MIA) techniques.

Abstract

A major focus of the series of studies was to evaluate the ability of the project’s thermal imaging technique to recover the defaced serial number through “blind” or “black box” studies. In these studies, the location of the numbers was largely unknown. This differs from all the samples used in the 2013 studies, since the location of the defaced numbers was known exactly or to a high degree of certainty. To achieve the “black-box” studies, the research team participated in the CTS Forensic Services serial number recovery test in July 2019 (Test No. 19-5251) and also in “black-box” studies with collaborators at the Idaho State Police Forensic Services Lab. The following results are reported: 1) the project team was unable to identify the six defaced serial numbers correctly when it participated in the CTS Forensic Services studies; 2) in the team’s blind studies that included subsequent recovery  by the Idaho State Police, the research team identified 8 out of 42 characters correctly, and the police lab identified 29 out of 42 characters correctly. From the ”Black Box” studies, dividing the number of characters identified correctly by the total number of defaced characters from all the samples, the success rate in correctly identifying a defaced serial number by thermal imaging is about 19 percent if the exact location of the number is unknown. If the location is known, based on the results from the project’s last grant, the success rate is likely somewhat higher. 77 figures and 6 tables

Date Published: January 1, 2022