A New Approach for Firearm Identification with Hierarchical Neural Networks Based on Cartridge Case Images

When a gun is fired, characteristic markings on the cartridge and projectile of a bullet are produced. Over thirty different features can be distinguished from observing these marks, which in combination produce a "fingerprint" for identification of a firearm. In this paper, through the use of hierarchial neural networks a firearm identification system based on cartridge case images is proposed. We focus on the cartridge case identification of rim-fire mechanism. Experiments show that the model proposed has high performance and robustness by integrating two levels self-organizing feature map (SOFM) neural networks and the decision-making strategy. This model will also make a significant contribution towards the further processing, such as the more efficient and precise identification of cartridge cases by combination with more characteristics on cartridge cases images