Technology Australia

Technology to spot stolen cars in traffic

A computer program that may enable moving police cars to automatically detect stolen or unregistered cars in surrounding traffic has been developed by researchers at UTS.

Developed with support from the ARC Discovery Projects scheme, the program uses new techniques based on hexagonal pixels, rather than the conventional square pixels.

The new techniques enable a computer linked to a camera to accurately identify and read number plates in real-time, meaning the numbers can be checked immediately against databases.

The UTS team, alongside its commercialisation partner UniQuest, is exploring a variety of possible industry partners and licensing opportunities for the program.

For example, recognising the program’s potential for detecting the plates of stolen cars, the team has consulted with the National Motor Vehicle Theft Reduction Council about its requirements.

Other applications being investigated include using the program to manage evasion of car parking fees, and as a real-time warning system for oversize and overweight trucks.

The research was carried out over three years (2004–2007) with ARC funding of $156,000. The grant enabled UTS to recruit research associate Dr Huaifeng Zhang from the Chinese Academy of Science, to assist with the project.

The project resulted in significant further development of Spiral Architecture, a powerful new data structure in which images are represented as collections of hexagonal pixels, in contrast with the traditional square pixels.

Project leader Professor Xiangjian (Sean) He from the UTS School of Computing and Communications said hexagonal pixels gave much smoother edges in images than square pixels, which tend to look like jagged steps when the image is highly magnified.

“The beauty of hexagonal pixels is that they can provide equivalent picture quality using 13 percent fewer pixels,” Professor He said.

“It’s not a new idea, but what our team has done is to use hexagonal pixels to develop much better methods of curve detection than is possible with square pixels, and this has opened the way for much quicker and more accurate shape identification.

“We are now world leaders in hexagonal pixel technology, and the potential is enormous—it could for instance provide improved resolution for still and moving digital cameras and could find many applications improving the object recognition capabilities of robots.”