Classification of fired cartridge cases using 3D image capture and a comparison of database correlation method performance


Kara I., Karatatar A.

JOURNAL OF FORENSIC SCIENCES, vol.67, no.5, pp.1998-2008, 2022 (SCI-Expanded) identifier identifier identifier

  • Publication Type: Article / Article
  • Volume: 67 Issue: 5
  • Publication Date: 2022
  • Doi Number: 10.1111/1556-4029.15089
  • Journal Name: JOURNAL OF FORENSIC SCIENCES
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Periodicals Index Online, Aerospace Database, Analytical Abstracts, BIOSIS, CAB Abstracts, Communication Abstracts, Criminal Justice Abstracts, EBSCO Legal Source, EMBASE, MEDLINE, Metadex, Veterinary Science Database, DIALNET, Civil Engineering Abstracts
  • Page Numbers: pp.1998-2008
  • Keywords: ballistic evidence, digital forensics, digital image analysis, image similarity measurement, matching performance, structural similarity, IDENTIFICATION, BULLETS, SYSTEM
  • Hacettepe University Affiliated: Yes

Abstract

When firearms are used, they leave unique marks on the fired bullets and shells. Examining these marks gives clues about the crime scene and the weapon of the crime. However, due to the increase in the number of incidents in which firearms are used, the increase in the amount of evidence makes ballistic investigations very difficult and prolongs the analysis time of the evidence. With the automatic image analysis and identification system, ballistic evidence can be examined very quickly and used to classify possible matching evidences. In this study, we present an approach based on Image Similarity Measurement for forensic examination of cartridges from automatic pistols. For this purpose, we used 500 images of 9 x 19 mm and 500 images of 7.65 x 17 mm cartridge cases obtained from 20 different brands and models of pistols. We divided the images of bullet cartridges into four different categories as breech face, firing pin, ejector mark, and combined evaluation, and we obtained these images from the BALISTIKA 2010 system. We investigated the classification and matching performance of the obtained four category images using 4 different methods: Structural Similarity (SSIM), Root Mean Square Error (RMSE), Peak Signal to Noise Ratio (PSNR), and Universal Image Quality Index (UQI). The results show that our Structural Similarity (SSIM) approach is effective in classifying and matching ballistic evidence.