Fernando Puente León
SPIE's Internat. Symposium on Intelligent Systems & Advanced Manufacturing, Boston, 1999.
This paper deals with an important task within forensic science - the automatic comparison of bullets for the purpose of firearm identification. Bullets bear groove-shaped marks that can be thought of as a kind of "fingerprint" of the firearm on their circumferential surface. To accomplish the comparison task, mainly the fine grooves on the bullet surface are of interest.
The presented approach is based on an automatic extraction of a "signature" describing the relevant marks. To enable a reliable feature extraction, high-quality images of the bullets are obtained. After a preprocessing step, a model-based abstraction is accomplished by adaptively projecting the image intensities of relevant grooves along their course. The resulting one-dimensional signals are not only very compact, they also have proven to provide a faithful representation of the surface information originating from the rifling of the firearm. However, an additional signal processing step is needed to separate those signal components describing the system parameters from the interesting individual marks. The performance of the methodology presented is demonstrated and quantitatively assessed with an image database of real bullets. It is shown that with our methods, the efficiency of an automatic identification of firearms can be dramatically increased.