Over the last year, the nonprofit investigative newsroom ProPublica has published a series of articles that called into question whether some forensic techniques produce reliable results. One of the techniques questioned is the matching of wear marks along the seams of blue jeans. Now, a leading forensic image analyst and a postdoctoral researcher have published a study finding that the technique produces limited evidence at best.
Along the vertical seams of any pair of blue jeans, wear marks gradually develop. The FBI has been using these light and dark patches as essentially a barcode that they have asserted is unique to every pair. The FBI uses this “wear mark barcode” to match pairs of jeans shown in surveillance photos or videos to those owned by suspects.
Wear mark analysis is one form of a technique called “image analysis” or “pattern matching analysis,” which has little scientific basis. ProPublica has been able to show that many defendants have been tied to crime scenes using image analysis. Sometimes, analysts have testified using baseless statistics about how likely it is that the images match.
The new study found two major flaws in blue jean wear mark analysis:
- The worn seams on different blue jeans are not unique like a bar code but instead appear extremely similar in many cases
- Taking multiple photos of the same pant seam under varying conditions can make the wear marks appear starkly different from one another
Considering those flaws together, the researchers concluded that “identification based on denim jeans should be used with extreme caution, if at all.”
The researchers studied images of jeans and used signal analysis to digitally convert the wear marks into numeric values to give an objective view of how different the jeans were from one another.
They had some success visually marking features like the FBI does. However, when the images were compared by computers, the computers struggled to identify photos of the same pair of jeans. The only way the researchers could match images in most cases was by allowing for a high rate of false positive results.
Ultimately, they found the technique helpful only about 30% of the time – and that is probably a best-case scenario. The researchers were using high-resolution photos of jeans that were known to be the same. In the real world, the evidence is usually low-resolution imagery from security cameras. Therefore, the technique may be even less reliable under actual conditions.
The researchers argue that all types of pattern matching analysis need to undergo scientific validation tests performed by independent laboratories.
“Mistakes in these identifications are costly, resulting in an innocent person being accused or sentenced and a guilty person walking free.”