Accession Number : ADA299050
Title : Firearms Identification Using Pattern Analysis and Computational Modeling.
Descriptive Note : Trident scholar project rept.,
Corporate Author : NAVAL ACADEMY ANNAPOLIS MD
Personal Author(s) : Jiong, Foo S.
PDF Url : ADA299050
Report Date : 09 MAY 1995
Pagination or Media Count : 73
Abstract : This paper describes a computational approach to the analysis of firearms forensic images, in particular, bullet images. Expended bullets usually bear characteristic surface markings associated with the mechanisms of the firearm that fired them. The traditional method of comparing bullet and cartridge markings involves the labor-intensive process of visual identification through a comparison microscope. This project proposes that the same process can be duplicated, and accelerated, using a low-cost video imaging and image processing system. The development of this project requires three important phases - "collection", "processing", and "analysis". In a laboratory scenario, the "barcodes" of the crime bullet and test bullet images are matched against each other. The barcode of the crime bullet is also matched against the barcodes of other potentially similar bullet images from the image database. A probability of match is generated for each comparison, and bullet images with a low probability of match are ignored. This effectively reduces the firearms examiner's workload. The bullet images with a high probability of match are retrieved, and displayed simultaneously to confirm the match.
Descriptors : *IDENTIFICATION, *SMALL ARMS, TEST AND EVALUATION, DATA BASES, MATHEMATICAL MODELS, IMAGE PROCESSING, SCENARIOS, COMPUTATIONS, LOW COSTS, SMALL ARMS AMMUNITION, COMPARISON, PROBABILITY, SURFACES, IMAGES, PATTERNS, LABORATORIES, MARKERS, GUNS, MICROSCOPES, VIDEO SIGNALS, LOW RATE, LITIGATION, CRIMES, CARTRIDGES.
Subject Categories : Cybernetics
Miscellaneous Detection and Detectors
Distribution Statement : APPROVED FOR PUBLIC RELEASE