This is a report on the examination of the impression markings on cartridge cases fired from semi-automatic pistols in order to determine the extent to which these markings can identify the particular gun that fired them, as well as whether such an identification can be quantified in terms of the likelihood that a match occurred by random chance, so as to determine whether improvements can increase the viability of a national database.
The study concludes that although cross-correlation algorithms may be unable to distinguish between large numbers of confocal cartridge case images, this does not necessarily reflect on the validity of the techniques of comparison microscopy. Differences between the cross-correlation coefficients from matching and non-matching cartridge cases can be increased by focusing on particular sized regions of correspondence. In addition, other approaches, such as pattern recognition, can also be used effectively in supplementing the cross-correlation techniques. Thus, by modifying the algorithms that determine the similarities, it should be possible to increase the number of cartridge cases that can be added to a database before it will be overwhelmed by false positives. The work reported was able to demonstrate that the size of the individual regions of corresponding topography on the breech faces of cartridge cases fired from the same slides were consistently larger than those that were consecutively manufactured, and also that there are other aspects to the matching that are not reflected by cross-correlation analysis. 33 figures and 12 references
810 Seventh Street NW, Washington, DC 20531, United States
Box 6000, Rockville, MD 20849-6000, United States