Generally, my assumptions of the ins and outs of crime scene forensics must be somewhat mythical. Perhaps my doubt is fueled by daytime television’s obsession with crime scene investigation units. The detectives sprawl out some high-tech gear and sexily perform some feats that no lab has ever seen, all for the sake of solving a little murder mystery via some dude’s encrusted drops of semen on a car seat.
I’d only expect that investigation units have access to some awesome modes of identity recovery, data analytics, predictive modeling, and the like. That’s why I felt a little surprised by the fact that only recently did research surface that demonstrates how law enforcement really ought to be catching criminals in their tracks—literally.
At the University at Buffalo, computer scientists led by Yi Tang have published new research (PDF) showing a novel computer system for identifying shoe prints at crime scenes. With several thousand footwear styles out there, there’s a clear need for deploying new technology that can sensitively match criminals with their sneakers. “[F]ootwear marks are found more often than fingerprints, and yet left largely unused due to lack of efficient and reliable tools,” the authors wrote.
But the hard part is identifying how to tell a computer to find matches for shoe prints. According to the research, 91.8% of the shoe prints looked can be fetched by tracing three simple components of their soles: “line segments, circles and ellipses.” In essence, the authors treat shoe prints like fingerprints, and have found a way to calculate similarities between different types of shoe.
Sample of shoe prints
Although footprints being used as evidence isn’t anything new, the restrictions as to what constitutes an acceptable “untainted” print for evidence sake has left a lot of scuffs and partial prints outside the courtroom. Using sophisticated modeling that can match partial sole prints to a database, the team hopes to make partial prints more valid.
The above photo shows a trio of partial prints being compared to a reference photo. The system breaks sole patterns down into shapes; lines are represented by green, circles in red, and ellipses in blue.
This image shows how scans of prints in the fields are modeled and then assigned to specific sole types. As with arches, loops, and whorls in fingerprints, shoes can be classified by their basic type to narrow down a match.
Using Footwear Print Distance or FPD, the research team looks computing the shoe prints at a crime scene in a similar way finger prints are examined. Referencing the distribution of curves and lines in relation to one another illuminates that shoe prints have quite unique characteristics from pair to pair, and should be highly considered by forensic investigations more often. This type of sole-chasing could help solve future crimes more efficiently.
I just can’t wait until it’s a cultural reality. That when we watch a movie in which criminals aren’t wearing bags over their feet, they’re are heckled just like the amateurs that show up gloveless at a bank robbery.