Graffiti isn't always used to spruce up abandoned homes, freeway overpasses or trains. Sometimes it's used by gangs to mark their territory or communicate about illicit activities. So this new system that accurately catalogues graffiti, and who's behind it, gives police a useful tool for identifying and staying one step ahead of local gangs.
Their traditional methods involve identifying the graffiti based on keywords, like the name of the gang or the artist, but that can often make searches ineffective and tedious. So Anil Jain, a computer scientist at Michigan State University, has developed an algorithm that compares the actual photos of the graffiti, looking for similarities in the artwork and design. Since they're all drawn by hand, the artists usually have a distinct style they repeat, whether they realise it or not, and this lets Anil's system identify not only the gang, but often the person holding the spraypaint can as well.
To further improve the accuracy of the algorithm, the images are sometimes outsourced to Amazon's Mechanical Turk service where humans are used to decipher any text. And when combined with the image matching techniques, the system has correctly identified the culprit about 65 per cent of the time. Not exactly perfect in a country where you're innocent until proven guilty, but the police seem more interested in using it as a tool to track gang activity, not necessarily for arresting individuals for vandalism. [New Scientist via Ubergizmo]