Application of Machine Learning to Toolmarks: Statistically Based Methods for Impression Pattern Comparisons

The authors propose to develop standardized methodologies to study and critically evaluate impression evidence. Three-dimensional vector data of the cartridge case impressions will be collected using high resolution 3D laser scanning and stored in a large database. This will generate a digital 3-dimensional map of the impression surface using X, Y, and Z coordinates. The data will then be examined with multiple multivariate statistical methods in order to quantify the criteria necessary to match a physical marking. ca/ncf