On the Study of Automated Identification of Firearms Through Associated Striations

Abstract

This paper presents possible application of image processing for automatic segmentation of striations and application of machine learning technique for pattern recognition and interpretation of those bullet striations. For identifying the bullet striations an automated technique is proposed. Experimentation was performed on four real test bullets of 5.56 mm calibre fired from two different firearms and two real bullets of 9 mm calibre fired from the same firearm. Experimental results show that it is possible to detect striations from microscopic images of bullets using image processing techniques. Shape features of these striations are computed whose patterns are analysed. Using K-nearest neighbour machine learning technique, we were able to classify firearms correctly with an average accuracy rate of around 84.67%.