Researcher outlines plan to digitize shoeprint evidence and build statistical foundations for footwear identification

An unnamed presenter at a forensic research session outlined plans to expand digital databases and statistical methods for footwear evidence, saying the field currently relies too heavily on subjective examiner judgments.

The presenter, who said he has about 30 years of casework experience, told attendees that panels such as PCAST and technical reviews show ‘‘it is impossible to assess the number of characteristics that must match in order to have any particular degree of confidence,’’ and argued that large, digitized datasets are required so comparisons can be made mathematically rather than by unaided human judgment.

The presentation described experiments dating back to the early 2000s that sought to test physical-match claims. In one early test the team digitally modeled torn silicone contours and challenged experts with intentionally confusing comparisons. The speaker said those tests demonstrated limits to unaided human comparison and motivated a computational approach: "So what we have to do is run this computer system about a small fragment ... against all the 40 centimeters of the other rim to find can we find what is the minimum length of physical match that we can identify. We reach 2 and a half millimeters which is 1 tenth of an inch and we found that when you go so lower distances then you get a material based behavior."

The presenter described a semi-automatic workflow presented in the talk as "CESA": human operators mark impression edges on scanned test impressions, software extracts border, orientation and location, and accepted impressions are stored as digital ‘‘signatures’’ in a searchable database. He said scanning was done at 600 DPI (with 1,000 DPI also used in research) and that an operational database used in demonstrations contained roughly 13,000 individual mark records ("accidentals").

He emphasized limitations in earlier work: prior analyses were based on about 400 shoes collected from real cases with uncontrolled wear and varied contact areas, which complicates black-box testing and dependency analysis. To address that, the presenter described a new, multi-site research collaboration with CSAFE (Center for Statistics and Applications in Forensic Evidence), Iowa State University and a Hebrew University laboratory. The project will collect repeated impressions from 160 volunteers (two shoe sizes, 8 and 10) at Iowa State with impressions taken monthly and will digitize and analyze changes in marks over time. A small pilot in Israel gave 10 shoes to police officers and has been running for about 90 days, the presenter said.

The presenter framed the goal as making footwear evidence "mathematically based" and more useful in court, but he cautioned that translating laboratory and pilot results into courtroom evidence remains a pending challenge: "We have a small problem now taking our results ... and to translate them to evidence in court. We don't do it yet."

The talk concluded with a call to expand lab and field databases, identify stable features across wear and time, and develop statistical methods (likelihood ratios and other separation criteria) that can provide quantified measures of similarity and error rates to support forensic testimony.