Award Information
Description of original award (Fiscal Year 2017, $598,315)
As submitted by the proposer:
Given a questioned impression (crime scene) and a test impression (obtained from a shoe of interest), an important question to triers of fact is “did the shoe corresponding to the test impression also make the crime scene impression?” The 2009 National Academy of Sciences report and 2016 President's Council of Advisors on Science and Technology report state that forensic footwear identifications are largely subjective. This has raised questions about the reliability and scientific validity of footwear examinations. Both reports have expressed the need for quantitative assessments of footwear evidence using scientifically valid methods.
Our proposed research will develop quantitative, objective methods for footwear impression comparisons. These methods will be used by footwear examiners not only to provide more scientifically valid information for court cases, but also to improve intelligence gathering through database searches. The methods will have a higher degree of repeatability and reproducibility than today’s methods, with quantified estimates of accuracy.
The project will develop quantitative methods for identifying features present in questioned and known footwear impressions, and metrics for summarizing the degree of correspondences and discrepancies among these features in the form of comparison scores. This will assist footwear examiners by providing them with the comparison scores and other quantitative information to help place the scores in context. A key challenge is the difficulty in accurately extracting features corresponding to footwear-related structure from crime scene impressions that are often partial, noisy, and of low quality. Our research plan is to use a hybrid human/automated feature extraction step, followed by an automated quantification of the degree of correspondences and discrepancies between the impressions being compared.
The three major tasks in the project are (1) develop algorithms for hybrid feature extraction based on a human examiner guiding a computer algorithm that automatically extracts features, (2) collect a modest library of footwear data of known ‘mated’ and ‘non-mated’ pairs of impressions for development and testing, and (3) develop algorithms for computing comparison scores that summarize the degree of correspondences and discrepancies of features in two impression images. These tasks will be implemented over a two-year time frame.
NIST is partnering with footwear examiners from the FBI, Charlotte County (Florida) Sheriff’s Office, and the Defense Forensic Science Center.
Expected products include, in addition to conference presentations, papers detailing our scientific results to be submitted to major forensic journals, and a technological prototype consisting of computer code that implements our algorithms.