|
Activity Number:
|
667
- Statistics, Science, and Society
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 2, 2018 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
IMS
|
|
Abstract #329910
|
Presentation
|
|
Title:
|
Methods for Automatic Groove Identification in 3D Bullet Land Scans
|
|
Author(s):
|
Kiegan Rice* and Heike Hofmann and Ulrike Genschel
|
|
Companies:
|
and Iowa State University and Iowa State University
|
|
Keywords:
|
forensic science;
bullet;
3D
|
|
Abstract:
|
Following the 2016 PCAST report on the validity of feature-comparison methods in forensic science, research focus has shifted towards automated comparisons of bullet marks. One avenue for automated comparisons is based on high-resolution 3D scans of bullet lands. Statistical learning techniques are used to compare scans and quantify the strength and quality of matches between bullets. In order to fully automate the quantification process an algorithm has to be able to automatically detect the location of the edges of the lands (grooves). Incorrect identification of groove locations is shown to lead to misidentification of key characteristics during the automated process and thus a significant increase in overall error rates down stream in the algorithm. Groove location is an inherently inverse statistical problem, as the relative heights of the bullet land are needed to determine the groove locations. Current solutions that do not address this issue are susceptible to numeric instabilities. We are proposing an approach based on robust linear models to provide solutions that are more reliable and therefore lend themselves better to automation without the need for human intervention.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2018 program
|
