Algorithm for Automatic 3-D Scanned Bullet Identification Based on the Principle of Human Vision

Abstract

Bullet identification is a complicated work which demands high accuracy rate. In this paper, we try to find a reliable and fast automatic bullet identification algorithm using image processing technology. The key and most difficult step is to find an image feature which can describe the unique striation pattern on the bullet. When we do bullet identification by human-eye, we try to best match the striations of the bullets. It’s mainly according to the order of the striations and the distance between them. Based on this principle, a new algorithm for automatic bullet identification is proposed. Using the bullet’s visual graphic set which is offered by the Three-Dimensional Laser Color Scanner (3DLCS), we firstly do pre-processing to the bullet unwrapped image, such as image enhancement, edge detecting, binarization, thinning and denoising, to obtain an image with clear striations and low noises. Then we find an effective image feature based on the principle of human vision: the bullet feature vectors (BFVs). According to actual needs, we give a reasonable definition for vector distance (VD) between vectors of different dimensions, and take VD of BFVs as a measurement for bullet similarity. Experimental results prove that our algorithm can correctly find the bullets with high trace similarity and do bullet identification successfully.