Image search firms target police, marketing users

Winnipeg law enforcement will start using software that automatically applies biometric analysis to identify faces of child pornography victims and offenders, in an effort to both hasten the process and minimize the emotional trauma the gruesome task inflicts on police officers.

LACE (Law Enforcement Against Child Exploitation) software, created by Quebec-based BlueBear LES Inc., is based on three algorithms. The first, soft image matching, searches for matches across pictures that are mathematically or numerically different, like .gif and .jpeg, thumbnail and large resolution image, and cropped and original-sized picture.

Another algorithm, face extractor, can extract faces from image and video sources by identifying features like the nose, eyes and mouth. Finally, the biometric facial algorithm identifies faces and matches them against known faces in other sources.

According to the company’s CEO, Antoine Normand, the software has a 100 per cent accuracy rate with no false positives when the trigger, or confidence level, is set at 80 per cent. At that level, although the technology may fail to make certain matches or identifications due to, say, a partially deleted file, Normand assures there are never false positives.

While LACE has applications beyond law enforcement like radiology to detect breast cancer growths in before and after images as well as some military applications, BlueBear “decided to focus on the security world in general… because the child pornography problem is a huge problem,” said Normand.

For release in the near future, Normand said BlueBear is currently working on expanding its soft image matching technology to video content “to be able to tell you that this scene in this AVI document was a subset of that larger MPEG document that was seen.” He anticipates software testing will begin in the fall or winter of 2009 with a possible release at the end of that period.

However, in the longer term, BlueBear is focusing its biometric technology on counter-terrorism to facilitate recognition in “an almost real-time face extraction,” said Normand, and extract faces from live video for comparison against a database of known faces. That software is still at the drawing board and will eventually be tailored to meet specific requirements of individual law enforcement agencies.

Another vendor of image identification and visual search software, Toronto-based Idee Inc., is in the midst of conversations with law enforcement agencies in Canada and the U.S. around its TinEye technology, a large scale image search engine that pinpoints in real-time where an image has appeared online.

Currently the TinEye database houses 700 million images (scheduled to hit one billion in a couple of weeks) against which users can compare an uploaded image and retrieve matching images.

Leila Boujnane, CEO and co-founder of Idee, said that as useful as image identification technologies can be, the key to real success is integrating the best of different technologies like face identification and image background analysis, instead of relying on one a single approach.

In the case of image background analysis, said Boujnane, very often law enforcement needs to “focus on the background because that will give you clues as to where this is taking place, have you seen it before, can you correlate to other images that have given you information about a location.”

It’s therefore critical to understand how to select the technology strengths you want to combine, and, she said, “piecing together these five or six or ten highly-specialized technologies to be able to have a system that allows you to do the work you have to do.”

Boujnane, too, can name numerous applications for image identification technology besides fighting crime. Business executives use TinEye to track where their images have appeared online. And, it can be “part of a holistic marketing approach” for the professional or amateur photographer of those images who may want to know where, and how many times, they get published.

Image identification can also be combined with the ubiquity of the mobile device to assist today’s consumer. “You can literally hold your iPhone to a book or a CD cover and take an image and if you want additional information about that, all you need is a system that takes that image, enlarges it, and then compares it against a large image collection,” said Boujnane.

The days of software and hardware limitations are gone, and for the first time, said Boujnane, “capabilities of the technologies and of the hardware are both merging together to give us an incredible ability to now deliver these types of services in a large scale.”

Related Download
Improving the State of Affairs With Analytics Sponsor: SAS
Improving the State of Affairs With Analytics
Download this case study-rich white paper to learn why data management and analytics are so crucial in the public sector, and how to put it to work in your organization.
Register Now