1. The data used was 3D topographical images of each bullet. Each image is in x3p format, which is an array of surface measurements at the micrometer level. R was used to streamline formatting.

  2. The traditional method of examining bullet striations involves placing the bullets under a comparison microscope and manually aligning them to match striations. Match thresholds have been established as “standard practice”, which is a less statistical method of examination. Instead, Hare et al suggest creating and using a 3D reference bullet database that computes distributional difference between known matches and known non-matches. The reference database was used to derive features and train a statistical model.

  3. They used random forest to provide an estimation of uncertainty (Hare, Hofmann, & Carriquiry, 2016, p. 23).

  4. Cross-correlation is the measure of similarity as a function of the displacement. It is commonly used to analyze multiple time series (Peterson, 2001). This method was used in this study for pattern recognition. In statistics, this function refers to the correlations between the vectors and entries of x and y (Cross-correlation, 2017).  

References Cross-correlation. (2017, June 23). Retrieved from Wikipedia: https://en.wikipedia.org/wiki/Cross-correlation

Hare, E., Hofmann, H., & Carriquiry, A. (2016). Automatic Matching of Bullet Land Impressions. The Annals of Applied Statistics.

Peterson, B. (2001, October 1). CROSS-CORRELATION ANALYSIS. Retrieved from Variability of Active Galactic Nuclei: https://ned.ipac.caltech.edu/level5/Sept01/Peterson2/Peter4.html