U.S. flag

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

Ballistics Matching Using 3D Images of Bullets and Cartridge Cases: Project Summary

NCJ Number
182566
Author(s)
Benjamin Bachrach
Date Published
1999
Length
11 pages
Annotation
This report presents relevant background information and describes the method used to determine the feasibility of using three-dimensional (3D) information from a bullet's surface to improve the matching rate of existing automated search and retrieval systems.
Abstract
Under two-dimensional (2D) analysis, the captured data for a bullet yield a "photograph" of the surface of the specimen. Algorithms developed to correlate different specimens based on 2D captured data have provided satisfactory results in the case of cartridge cases, but disappointing results in the case of bullets. The current project was launched to determine whether there are advantages to the use of 3D captured data compared to 2D captured data. This involves using a depth measurement of the bullet's surface instead of using a "photograph" of the bullet's surface under 2D captured data. The 3D normalized data set is the result of mathematically processing the digitized data to remove all systematic errors introduced during the capture process. Once the data are normalized, the most significant features of the bullet emerge clearly. To determine the feasibility of using 3D information from a bullet's surface to improve the matching rate of existing automated search and retrieval systems, the elements of an acquisition component were developed and implemented. Together with the acquisition component, a preliminary version of a correlation component was developed to verify the usefulness of the 3D captured data. The complete automated search and retrieval system was tested through a number of independent evaluations. Among these evaluations, a number of so-called "blind tests" were performed. The task was to identify which gun fired each of the questioned bullets based on the data obtained from control bullets. In all cases the system was able to perform in a satisfactory manner, making few mistakes in the identification of which gun fired each of the questioned bullets. 7 exhibits

Date Published: January 1, 1999