License plate recognition system leverages megapixel cameras, machine learning, artificial intelligence, and rugged computing capabilities for faster, more effective, and more efficient enforcement.
In cities with vibrant downtowns, parking enforcement plays an important role in the success of local merchants. Time limits for street parking encourage patron turnover, which boosts business revenue. Parking enforcement has historically involved sending officers out on foot, armed with ticket pads and chalk to mark the tires of parked cars. However, this is not the most efficient way to realize the benefits of parking time limits.
Taking an Automated Approach
Recent developments in cameras, sensors, and processors — combined with court rulings questioning the legality of marking tires — have hastened the need for computerized systems to help enforce parking limits. With such systems, officers drive their routes in vehicles outfitted with cameras and computers that can automatically identify parking violations.
Tannery Creek’s autoChalk systems employ computers, megapixel color cameras, precision GPS, optional laser detection, and sophisticated software to scan parked vehicles and their license plates to determine if they are parked legally. The systems can scan two vehicles per second at speeds up to 60 kilometers per hours (37 mph). autoChalk uses a combination of automatic license plate recognition (ALPR) and vehicle recognition technology to accurately determine the plate number, color, contour, and length of each vehicle.
“Our autoChalk systems can make parking enforcement officers as much as five times more productive,” said Bill Franklin, a founder and the president of Tannery Creek Systems. “The officer scans at 30 miles per hour, and the system automatically alerts them when it detects a potential infraction. The officer stops and examines the vehicle images on the computer screen and issues a ticket if appropriate.”
Because many vehicles have license plates that are either missing, covered, or not legible (in Tannery’s experience, typically 2 percent of vehicles do not have legible license plates), autoChalk optionally employs a tough and rugged laser that can measure the length of a parked vehicle to within about an inch. Using length combined with color analysis of parked cars, autoChalk’s proprietary and patented technology can identify parked vehicles that have not changed location from one scan to the next. In jurisdictions where only a rear plate is required, parkers often deliberately hide their license plates by backing into parking stalls. However, this and other methods do not fool autoChalk’s detection, which significantly improves the effectiveness of parking enforcement.
The system also employs precision GPS technology that narrows a vehicle’s location to within a radius of 2 meters (6.5 feet). Precise placement of the parked vehicle is crucial, as it must assess if the vehicle can park in that location or if it is the same vehicle previously seen in that location (important for time-limited parking, pay-by-space enforcement, and parking studies). Vehicles with a permit or a paid parking session show up with the letter P on the parking officer’s heads-up display. Violations are also flagged, so the parking officer can stop to issue citations. Automated scanning and audible alerts make for a safer heads-up enforcement system, reducing repetitive stress ailments, severe injury, and traffic accidents.
A System that Keeps Everyone Honest
An automated system is more efficient with regard to officers’ time. The data recorded, such as photos and GPS location, facilitates a transparent and more accurate enforcement process, making it much more difficult for parkers to contest legitimate citations. In short, high-quality images and real-time data keep both drivers and cities honest.
In cities with pay-by-plate systems, drivers enter their plate information and pay for parking via an app. Collecting and processing that data requires a complex system with many components working smoothly in concert. For example, a mobile parking enforcement vehicle needs reliable wireless access to cloud-based databases to match vehicles with parking transactions.
As the system captures images, the machine vision software gets to work. autoChalk uses machine intelligence to perform license plate recognition. The software, which runs in Linux, accurately reads plates at angles from 0 to 90 degrees and from all jurisdictions across the United States and Canada. The software also tolerates a wide range of image quality and in some cases partial obscuring of license plate characters by dirt or snow. Machine learning enables artificial intelligence algorithms to correctly identify new license plate formats and handled incorrect readings.
The System Needs a Brain
ALPR is very CPU- and graphics-intensive because it must look up images of millions of plates within a small fraction of a second to correctly identify the plate within the captured image. Hence the graphics processing capability of the processor is a critical part of the system’s machine intelligence, with the graphics processing unit (GPU) handling number crunching, correlation, and vector processing to perform license plate recognition.
It uses GPS, the cloud, and GPU capabilities to process data and images from the cameras, ascertain potential infractions, and send audible alerts to parking enforcement officers as they scan. Processing demands are huge given the rapid inflow of 40 photos per second and the handling of millions of images to get accurate license plate readings.
Given that it is used in vehicles across North America, it is crucial that GPU computers work well in harsh environmental conditions, such as temperatures below minus 40 or above 100 degrees Fahrenheit. In seaside towns, humidity can approach 95 percent, a difficulty exacerbated by airborne salt. Vibration and shock are constant hazards, along with challenges like power fluctuations, radiated electrical noise (electromagnetic interference or EMI), and a lack of ventilation or cooling. So, components and systems need to be resilient. Thread-locking compounds, glue, and potting are used to prevent electronic components from bouncing around, breaking, or failing.
“It is a challenge to do license plate recognition in a moving environment,” said Franklin. “The systems are constantly shaking, which can cause screws to fall off, and the external equipment has to survive exposure and operate in the elements 365 days a year.”
“The Toughbook has an optional GPU, but the CoastIPC GPU computer is some five or six times faster,” said Franklin. “We also looked closely at how the computer was put together, fastened, locked down, and how it handles vibration, extreme cold, and heat.”
“For us to offer a really reliable product to our clients, each component has to be more reliable than our product is as a whole,” said Franklin. “CoastIPC has worked closely with us to deliver the solution we need to build an efficient and dependable system.”
Jimmy Carroll is an account executive with TECH B2B Marketing. He can be reached at jimmy@techb2b.com