Three dimensional visualization and comparison of impressions on fired bullets

Currently, optical devices, such as microscopes and CCD cameras, are utilized for identification of bullets and tool marks in the field of forensic science. While these optical methods are easily manageable and effective, they are under great influence of illumination condition. In other words, appearances of striations through these optical devices have possibility to be changed by lighting condition. Besides these appearance-based approaches, we can utilize three dimensional (3D) geometric data of tool marks that are free from lighting condition. In this study, we focused on 3D geometric data of landmark impressions on fired bullets for identification. We obtained the 3D surface data of tool marks by a confocal microscope and reconstructed virtual impressions on a PC monitor from the geometric data. Furthermore, the 3D data are exploited to numerical matching of two surface shapes. We also visualized the difference of two shapes. In order to do this, two surface models are aligned automatically. In this process, pairings of correspondent points on both surfaces are determined. Distance analysis between these pairs leads to a shape comparison. Since comparison results are visualized, they are intuitive and easily perceptive.     

Estimation of bullet striation similarity using neural networks

A new method that searches for similar striation patterns using neural networks is described. Neural networks have been developed based on the human brain, which is good at pattern recognition. Therefore, neural networks would be expected to be effective in identifying striated toolmarks on bullets. The neural networks used in this study deal with binary signals derived from striation images. This signal plays a significant role in identification, because this signal is the key to the individually of the striations. The neural network searches a database for similar striations by means of these binary signals. The neural network used here is a multilayer network consisting of 96 neurons in the input layer, 15 neurons in the middle, and one neuron in the output layer. Two signals are inputted into the network and a score is estimated based on the similarity of these signals. For this purpose, the network is assigned to a previous learning. To initially test the validity of the procedure, the network identifies artificial patterns that are randomly produced on a personal computer. The results were acceptable and showed robustness for the deformation of patterns. Moreover, with ten unidentified bullets and ten database bullets, the network consistently was able to select the correct pair.