Student improves licence plate recognition technology

Juan Yepez has developed a new, faster and more accurate software to improve existing systems that can identify a car’s licence plate by simply analyzing plate images.

Plate recognition technology is becoming more common to enhance security, but current systems are expensive and relatively slow, said Yepez and his supervisor Seok-Bum Ko.

“Our goal is to create affordable software that can be used with any type of camera,” said Yepez, a master’s student in computer engineering. “It could be used for many applications such as automatically finding cars that run red lights, exceed speed limits or enter restricted areas.”

Yepez and Ko’s software uses an improved algorithm, a set of calculations performed by the computer. Their new software can “read” plates almost two and a half times faster than software currently on the market and would be cheaper because it doesn’t require an external image-processing server.

Yepez expects his software will bring down costs considerably. Commercial software costs around $1,000 for a single camera installation and routes the image-processing to servers that cost up to $10,000 for a license.

“Our research can help companies that currently work with licence plate recognition to update their software and develop new devices with the algorithm we designed,” said computer engineering professor Ko.

Yepez used a research database of licence plate images for lab testing. His system was able to “read” 100 per cent of all clear images of licence plates and more than 98 per cent of plates overall—including clear, dirty and blurred ones. This result has never achieved before in research using similar algorithms. 

“The main advantage of our software is that it doesn’t need any special equipment because commercial software need special cameras and a high-speed computer to identify the licence plates,” said Yepez. 

The software could make a difference in helping police track stolen cars. Yepez said with his software, police and parking enforcement can share licence plate information in real time. Since his system doesn’t need any image-processing servers, police would not require additional expensive devices to access parking enforcement servers.

Yepez said he knows how to custom tailor the new software for police needs, but added more research is needed before it hits the market. His next step is to test the software on actual cars in parking lots.

Yepez has always been interested in software for image recognition. He had worked on it in Ecuador, his home country, where he still runs his own business for hardware and system design.

He came to the U of S last year on a scholarship from the Ecuadorian government through CALDO, a consortium of Canadian universities focused on fostering international education.

“I chose Canada because I really believe that it is a good country to be with family and I wanted to do research at this university,” said Yepez. “Canada opened the door to a world of opportunities for me.”

Juan Yepez’s story is captured on YouTube.

 
Federica Giannelli is a graduate student intern in the U of S research profile and impact unit. 

This article first ran as part of the 2017 Young Innovators series, an initiative of the U of S Research Profile and Impact office in partnership with the Saskatoon StarPhoenix.