An efficient automatic firearm identification system

Abstract

Automatic firearm identification system (AFIS) is highly demanded in forensic ballistics to replace the traditional approach which uses comparison microscope and is relatively complex and time consuming. Thus, several AFIS have been developed for commercial and testing purposes. However, those AFIS are still unable to overcome some of the drawbacks of the traditional firearm identification approach. The goal of this study is to introduce another efficient and effective AFIS. A total of 747 firing pin impression images captured from five different pistols of same make and model are used to evaluate the proposed AFIS. It was demonstrated that the proposed AFIS is capable of producing firearm identification accuracy rate of over 95.0% with an execution time of less than 0.35 seconds per image.

A Comparative of Two-Dimensional Statistical Moment Invariants Features in Formulating an Automated Probabilistic Machine Learning Identification Algorithm for Forensic Application

“…Although the execution time for a single analysis can be significantly reduced from several weeks to several hours by utilising these market-available ballistics identification systems, analysing the specimens remains dependent on physical interpretation in order to authenticate the analytical results. In the past few decades, several semi-automated and automated feature-based probabilistic machine learning identification algorithms have been proposed to address this limitation [1][2][3][4][5][8][9][10][11][12][13][14][15]. Comparing these semi-automated and automated probabilistic machine learning identification algorithms to the market-available ballistics identification systems, the principal benefits of these probabilistic machine learning identification algorithms do not depend on physical interpretation and imposed a short execution time ranging from seconds to minutes.…”

“…Therefore, the first principal objective of this article is to develop an improvised automated machine learning identification algorithm for ballistics that utilises the unweighted least square-fitting circle to detect the position (anchor point, A) of the centre-firing pin impression circular boundary with radius r. In this research, the firing pin impression is the principal focus key impression in developing the improvised machine learning identification algorithm for ballistics. This is because the previous empirical studies conveyed the invariant properties of this impression in ballistics identification rather than others' key impressions on the fired cartridge cases [2][3][4][5][6][11][12][13][14][15].…”

“…Due to the limitation of the machine learning identification algorithm for the ballistics in the previous studies, Chuan et al [5,12,13] have proposed an improvised automated machine learning identification algorithm for the ballistics from the series of research works of Ghani et al [2,3,4] and Liong et al [20].…”

“…STEP 4: Otsu [34] threshold selection method is employed to binarise the image 𝐏 𝐍 . Otsu's method is employed in this study because this method is widely and successfully utilised in multidisciplinary applications [5,6,12,13,22,35,36,37] even though this threshold method has been introduced over four decades ago. An optimal threshold value, 𝑤 ̃, is attained by maximising the function defined in Eq.…”