Yesterday, Today, Tomorrow – License Plate Recognition

The LPR industry’s history is much longer than many people realize. It goes back to 1976 when it was invented at the Police Scientific Development Branch in Britain, initially for policing and security applications. The first commercial LPR systems began to be deployed in 1979 and the first mass deployment of LPR was the “Ring of Steel” for the City of London in 1993. Our predecessor company HTS, deployed one of the first LPR uses for parking management in Israel in 1995.

Credentialed vehicles will be able to enter and exit a facility in the most seamless/frictionless/quickest and accurate manner imaginable.

For approximately the next fifteen years, LPR technology continued to evolve, advance, and improve the recognition accuracy of the OCR generally performed using early neural net technology, which is the predecessor to today’s Artificial Intelligence (AI) revolution. Within the parking arena, most of the LPR deployments prior to 2010 were within the airport market and almost entirely in the Tier 1 airport locations or high-security access control. This was mainly due to cost per lane being much higher (2x+) than today, the volume of vehicles in inventory at the facility, cost per ticket or parker, and an increased volume and value of lost and switched tickets.

Approximately ten years ago, the technology evolution of utilizing much lower cost off the shelf CCTV cameras with rolling shutters versus global shutters for low-speed LPR along with IP video merged with computer vision algorithms. This drove commercial applications for LPR beyond the traditional policing, high security, and airport parking markets. Fast forward a decade and today, even with all the advances in technology, reduction in costs, and improvements in LPR accuracy, it is estimated that only about 10 percent of all paid parking lanes are deployed with LPR as either the main or secondary credential for ingress and egress. 

While this varies by state/city or market like airports, universities, municipalities there remains a tremendous opportunity for continued growth and expansion throughout the parking arena and further into many more core commercial types of garages and lots. Today, our organization has over 6,000 lanes of LPR deployed in the parking market, and although we also have deployments at 40 percent of the top 30 airports in the U.S. by passenger traffic, our growth and the industry are no longer reliant on the early adopters like airports for finding value in this technology. Most all LPR developers and manufacturers routinely achieve 95 percent and greater accuracy, and many times achieve 99 percent with the caveat of “machine-readable, non-exception license plates,” but we still are not a perfect 100 percent for all kinds of legitimate reasons.

While LPR fights for its position versus RFID, Proximity Readers, BLE, and app-based credentials, we envision a world where license plates remain relevant, accurate vehicle data is paramount, and the post-Covid demand for frictionless, ticketless customer experience is tantamount for success and repeat business. When compared to many of the credential options available now (all have limitations), LPR remains a tried and true technology that works not only for monthly/credential parkers but also for transients. It improves their experience with the facility, and the speed with which they can exit is incredibly quicker than with other credentials. 

Today, many of us in LPR development are utilizing the latest deep learning AI neural network models to train our LPR engines for optimum performance. However, the world is changing, and as our industry evolves most folks just expect LPR to work, do so reliably, and do it every time the same way. To that end, we continue to invest in providing loopless, gateless, gated, and free-flow based solutions with value-added vehicle data points beyond LPR. These include vehicle recognition of make, model, color, speed, direction, vehicle classification, and occupancy counting, to name a few. In the pre-Covid world, the major differentiators for LPR selection were “what is your accuracy and price?”

Tomorrow’s leaders in this space will certainly be close to the 100 percent holy grail of license plate reading accuracy, however, there will be a number of differentiators between what was traditional LPR and the newer trends of vehicle data-driven information. This includes the prior mention of vehicle make, model, color, direction, occupancy, classification, state jurisdiction, and, of course, the license plate credential. In marrying these data points together, even if one piece of the hypothetical string of “White, Mercedes, 350, ABC123, New York” is off by one component the relative certainty of a vehicle match is increased. This is inching us closer to the ultimate accuracy/performance variable. 

Dynamic pricing can be implemented based on vehicle classification of small car, large car, SUV or commercial vehicle. Credentialed vehicles will be able to enter and exit a facility in the most seamless/frictionless/quickest and accurate manner imaginable. It is always the last few percent of perfection that are difficult to achieve, and as an industry, we are at the cusp of that reality today. Also, just as important as the core technology chosen is the company you choose to keep company with. Look for an organization that is financially stable, owns the IP and patents for their technology, has the resources to move and fund the growing Hardware as a Service model for LPR, and has a team of people in support and development that you trust and can count on when an issue arises.

John Whiteman is an expert in License Plate Recognition for over ten years with HTS which is now omniQ VISION and currently is Executive Director of Sales-Parking and Mobility which is a division of omniQ. John can be reached at jwhiteman@omniq.com.

Leave a Reply

Your email address will not be published. Required fields are marked *