Optimal compression and binarization of signature profiles for automated bullet identification systems

In some automated bullet identification systems, the similarity of striation marks between different bullets is measured using the cross correlation function of the compressed signature profile extracted from a land impression. Inclusion of invalid areas weakly striated by barrel features may lead to sub-optimal extraction of the signature profile and subsequent deterioration of correlation results. In this paper, a method for locating striation marks and selecting valid correlation areas based on an edge detection technique is proposed for the optimal extraction of the compressed signature profiles. Experimental results from correlating 48 bullets fired from 12 gun barrels of 6 manufacturers have demonstrated a higher correct matching rate than the previous study results without correlation area selection processing. Furthermore, an attempt to convert a traditional profile with multiple z-quantization (or gray scale) levels into a binary profile is made for the purpose of reducing storage space and increasing correlation speed.