Enlarge /. The "barcode" pattern of light and dark dots along the seam of a pair of jeans.
Is every pair of jeans like no other? According to forensic FBI analysts, the patterns observed on denim are reliably unique and can be used to identify a suspect in surveillance material.
The problem is that this technique has never been rigorously tested and the evidence it provides may not be as strong as it was claimed. An article published in PNAS this week put the denim pattern analysis through its paces. It finds that pairing identical jeans is not particularly good – and may cause a number of false alarm errors.
For some time there have been rumors about the reliability and quality of commonly used forensic techniques. In 2009, the National Academy of Sciences published a weighty report stating that, apart from nuclear DNA analysis, “no forensic method has the ability consistently and with a high degree of certainty to demonstrate a link between evidence and evidence a particular person or source. "
The problems with forensic evidence – including fingerprint, blood stain, and ballistics analysis – have terrible ramifications for the real world. According to the National Registry of Exonerations, nearly a quarter of unlawful convictions in the United States in the past 30 years can be attributed to incorrect or misleading forensic evidence.
Computer scientists Sophie Nightingale and Hany Farid wanted to look at one technique in particular: photographic pattern analysis, in which the detailed patterns on faces, hands or clothes are compared between suspects and crime scene photos. For example, jeans have a “barcode” pattern of dark and light spots along their seams.
These patterns were used as central evidence for the sentencing of people. But is this kind of analysis reliable? That has not been established. To try it out, Nightingale and Farid went out to buy 100 pairs of jeans in thrift stores. They laid the jeans flat on a hard surface, photographed the seams along the legs and digitally traced the pattern of light and dark dots along the seams. To improve their sample, they had Amazon Turk employees provide pictures of another 111 couples that were photographed with careful instructions.
The researchers then began to quantify how different the patterns were for different jeans. Obviously, a lot of randomness plays a role here – two pairs could happen to be pretty similar, while two other pairs could happen to be completely different. And most couples would fall somewhere in the middle with some degree of similarity. Based on these measurements, Nightingale and Farid worked out the area of similarity between the "barcode" patterns on different jeans.
The important question, of course, is whether these patterns can be used to determine whether two pictures show the same jeans. The researchers selected 10 jeans and each took 10 photos with different cameras, with different lighting and with different drapes. What they found was that any given pair of photos could come back with many similarities, but also come back with very different readings on the pattern. The spectrum was broad – as Nightingale and Farid point out, soft material that is photographed in different ways will have distortions that vary from image to image.
So if a pair of jeans can look noticeably different in different photos, is denim pattern analysis really a useful forensic technique? The researchers used their measurements to estimate how often a true match would occur and how often their jeans would trigger a "false alarm" – a score that looked like a match, even though the images actually came from two different pairs.
They found that the false alarm rate could be as high as one in a thousand. Given that the FBI has done a photographic pattern analysis in hundreds of cases every year, this is a reasonable option. The actual match rate was also not particularly high at around 40 to 50 percent, depending on factors such as the length of the seam to be analyzed.
This means that the jeans fitting technique is likely to be a hit and miss – it won't find actual similarities most of the time and may trigger a high false alarm rate. And all this under controlled test conditions with high-quality pictures and jeans that are beautifully flat and not grainy, and non-grainy safety material that shows the wearing of jeans. On the other hand, various characteristics such as damage, branding and size could confirm an analysis to improve the evidence in one way or another.
More work is needed to determine whether jeans can be analyzed more reliably using additional functions – and whether other pattern analyzes – such as freckles on the face or patterns on other types of clothing – are similarly unreliable. But for the moment, Nightingale and Farid write: "Identification on jeans should be done with extreme caution, if at all."
PNAS, 2020. DOI: 10.1073 / pnas.1917222117 (About DOIs).