Integrated portal helps crime-fighters link faces and names

Police often capture suspects in a photo or on surveillance video, but an image alone isn’t enough to make an arrest; investigators need to put a name to the face.

More than 39,000 county, state and federal justice professionals were already using the Pennsylvania Justice Network (JNET) to conduct secure investigations in a Web-based environment. JNET is an integrated portal that provides authorized users with access to public safety and criminal justice information compiled by federal, state and local sources.

But JNET couldn’t help investigators who only had a photo. So the Commonwealth of Pennsylvania Justice Network developed the JNET Facial Recognition System (JFRS) to help those investigators identify individuals in photos or videos.

JFRS uses the distinct measurements of each person’s facial features to compare an image of an unknown individual with images stored in the Commonwealth Photo Imaging Network (CPIN), which has more than 3.5 million photos of individuals who have been arrested by police. The system creates a facial plate from each photo when it’s entered into CPIN, which provides a uniform platform with centralized quality control for capturing, storing and processing images.

The plate detects and measures facial dimensions, which are documented for use within JFRS. The system then uses two distinct algorithms to compare faces to ensure that investigators receive a thorough set of results.

JFRS is accessed through JNET, so it’s available to every law enforcement agency in Pennsylvania, regardless of its size or technological sophistication.

First deployed in 2006, JFRS has since been updated with features to improve its usability and accuracy. The system is now completely Web-based, so users can access the application from within JNET without needing to download any software as they had before. The new system also includes improved algorithms for more accurate image-matching. It now has 3D facial-comparison tools that allow users to generate 3D models of both suspect and CPIN images. New tools also allow investigators to enhance poor-quality images for use in JFRS, and a video extraction tool enables users to upload images from video footage.

JNET plans to complete its training of investigators in the use of the new JFRS platform by mid-2012. But the system has already proved its worth since it was first introduced; investigators have used it to solve numerous cases, including homicides, robberies, and fraud and identity-theft cases.

Read more about the 2012 Computerworld Honors Laureates.