U.S. flag

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

Improving a Database to Help Identify a Vehicle by Using Paint Fragments

Identifying a car, truck, or other vehicle that has been involved in a hit-and-run accident often depends on examining paint fragments left at the scene by the vehicle.
Date Published
May 15, 2017

For years, investigators have relied on the Paint Data Query database, developed by the Royal Canadian Mounted Police, to identify the make of a vehicle by matching the physical attributes, chemical composition, and infrared spectrum of the paint, primers, and clear coating layers. The Paint Data Query database contains more than 21,000 automotive paint samples that correspond to more than 84,000 individual paint layers used on most domestic and foreign vehicles sold in North America.

Although the Paint Data Query database is extensive, it uses text to code the chemistry of each layer of a paint sample (typically two layers of primer, one of color, and a topcoat). The text coding is generic and leads to a large number of spurious hits that impair the accuracy of a search.

Another issue is that modern automotive paints have a thin color coat, which means microscopic fragments left at a crime scene may be too thin to obtain accurate chemical and topcoat color information. Forensic laboratories typically rely on PPG or DuPont color refinish books for making color comparisons on paint chips, and although those books are accurate on a macroscopic scale, they are less so on a microscopic scale.

Another concern with using the Paint Data Query database to identify a paint fragment is that many forensic laboratories use attenuated total reflection (ATR) spectroscopy for infrared analysis of automotive paints. Although ATR requires minimal sample preparation, the IR spectrum of an automotive paint sample obtained by ATR exhibits distortions and, as a result, hinders specific identification.

The researchers on this project addressed these problems in several ways. First, a search prefilter was developed to address the text coding issues in the Paint Data Query. The prefilter differentiates automotive paint samples by automobile manufacturer by examining the top clear coat and undercoat paint layers. Prefilters were developed that can identify the specific assembly plant based on the paint system used on the vehicle. The prefilters, developed for GMC and Chrysler vehicles, can identify the assembly plant or subplant where the paint was applied to the vehicle; and this, in turn, allows the model and line of the vehicle to be identified. The prefilters did not work as well for Ford vehicles.

A cross correlation library searching algorithm used in conjunction with the search prefilters outperformed OMNIC software currently used in many forensic labs for similar investigations, the researchers noted. “This is a potentially significant development as it can increase both the speed and accuracy of forensic automotive paint analysis,” they said.

They also developed a correction algorithm to allow ATR spectra to be matched using the IR transmission spectra of the Paint Data Query database. The correction algorithm is able to address distortion issues that currently limit the usefulness of ATR in matching automotive paint to the Paint Data Query database.

The overall goal of the project, according to the researchers, is “to enhance current approaches to data interpretation of forensic paint examinations and to aid in evidential significance assessment, both at the investigative lead stage and at the courtroom testimony stage.”

About this Article

The research described in the article was conducted under NIJ cooperative agreement 2012-DN-BX-K059 awarded to Oklahoma State University.

This article is based on the grant report “Improving the Paint Data Query Database to Enhance Investigative Lead Information From Automotive Paints” (pdf, 91 pages) by Barry K. Lavine, Collin White, Undugodage Perera, Koichi Nishikda, and Matthew Allen.

Date Published: May 15, 2017