Software links police sketches to mug shots

Michigan researchers have developed a face-recognition system that matches witness descriptions to database photos.

Here's a new spin on an old way to catch crooks--researchers at Michigan State University have developed algorithms and software that can match police sketches to thousands of mug shots in a database.

The research, by MSU computer science and engineering professor Anil Jain and doctoral student Brendan Klare, is being published in this month's edition of the journal IEEE Transactions on Pattern Analysis and Machine Intelligence.

Mug shots and composite sketches have long been used in police investigation, but witness descriptions can be inaccurate. The researchers say their project is the first large-scale experiment to link the sketches and mug shots, adding that results have been promising.

The algorithms compare eye, nose, and chin shapes, as well as other structures, in pairs of images. In the experiment, using a database of more than 10,000 images, the system found the correct person 45 percent of the time. The sketches were from real crimes in which the culprit was later identified.

"We improved significantly on one of the top commercial face-recognition systems," Klare was quoted as saying in a release.

The system is to be field-tested in about a year.

Featured Video