Multi-dimensional cluster analysis of class characteristics for ballistics specimen identification

The characteristic markings on the cartridge and projectile of a bullet fired from a gun can be recognised as a "fingerprint" for identification of the firearm. Over thirty different features within these markings can be distinguished, which in combination produce a "fingerprint" for identification of a firearm. The analysis of marks on cartridge casings and projectiles that have been fired, provides a precise tool for identifying the class of firearm from which a bullet was discharged. The measurement of features on ballistics cartridge cases and projectiles allow precise ballistics metrics to be obtained of cartridge case and projectile class characteristics for the identification of the make and model of the firearm. These different features within these markings and patterns can be identified, which in combination produce a "fingerprint" for any type of firearm. A variety of features can be observed under moderate or high magnification of the specimen. However, not all features are observable for all weapons under all conditions of discharge. This situation represents the classical situation of "missing data" when attempting to demonstrate "similarity" between specimens. This condition is overcome by seeking "feature by feature" comparisons without achieving identical matching for the crime scene and test specimens. This paper will describe progress in the development of a multi-dimensional cluster analysis model for forensic ballistics specimens to identify the type of weapon that produced these ballistics specimens. The cluster analysis will provide classification that is based on scalar shape and measurement parameters for the three-dimensional features of class characteristics. These class characteristics include calibre, firing pin mark, ejector mark, and extractor mark for the cartridge case; and number and widths of land and groove marks, and the direction of twist of rifling on the projectiles. The selection of appropriate class characteristics for cartridge and projectile can be mapped in N-dimensional space to provide clustering for particular weapon types. By mapping the crime scene specimen to the multidimensional ballistics data, the possibility of a match for identification may be achieved. This project has the potential to significantly improve the effectiveness and efficiency of tracing of firearms used in criminal activities