This is the first of two installments on Automatic License Plate Recognition systems by Jim Kennedy. He has been involved in the video industry since the mid-1970s, with a particular interest in closed-circuit video as it applies to traffic observation, and, over the past eight years, with vehicle license plate capture technology. (In the second installment, Kennedy will discuss the details of problems with mismatched license plate numbers.)
Just a few years ago, it appeared that the parking industry was ready to adopt the technology developed for toll roads whereby the license plate of a vehicle was captured by a camera and “read” through the use of some clever algorithms to which the image had been sent. The result was that the license plate number was now in a data format much as it would be if someone had typed it into the system and could, therefore, be associated with a transaction and/or ticket number.
For a variety of reasons, the technology has been slow to be incorporated into the revenue management side of parking systems, in spite of those early hopeful signals. The same was true of toll roads at one time, but once the value of the technology was proven, ALPR (Automatic License Plate Recognition) soon became an accepted piece of the toll violation process.
In one sense, there are no “violators” in parking, but there are those who would attempt to thwart the system. They are often referred to as “cheaters,” and those who successfully cheat the system can cost the parking management company a great deal of money.
There has been some debate lately over the actual size of the pay-for-space parking industry. The numbers range from about $11 billion to as much as $20 billion Let’s make it simple and say it is “only” $10 billion. If parking cheaters represent just 1 percent, it would result in losses equal to $100 million through such fraud. Even if it is only one-half of 1 percent, it is still $50 million. What if it is more? A 2 percent fraud would be $200 million. In any event, it is real money, no matter what the potential for parking fraud might prove to be.
ALPR is intended to minimize or even eliminate those losses by verifying that the car now at the exit is the same car as the one that took the ticket at entry. In short, the license plate is read at entry and the license plate number, along with the ticket/transaction number, is sent to a central database. When the car eventually arrives at the exit cashier, the license plate is again read and the license plate number is sent to the central database, where the associated ticket number is found. The license plate number just read at the exit is compared with the license plate number that was read at entry and which is associated with the ticket number. A matching algorithm compares both numbers and gives a notification to the exit cashier, or some remote supervisor’s station, indicating “match” or “no match.”
It is thought by some that with a very accurate ALPR system, tickets may become a thing of the past. The idea being that the entire entry and exit verification and fee calculation could be done from the license plate information. For a variety of reasons, the least of which is not the tactile record the ticket provides, a ticket-less system is probably still a bit into the future.
One of the more visible segments of the parking industry — and the one that has made ALPR technology a viable feature for aiding revenue management — is airport parking. In the past few years, several airports have incorporated ALPR as a means to discourage ticket swapping, lost-ticket fraud, cash leakage and, in general, a way to generate better revenue control and reporting. Airports such as Ben Gurion in Tel Aviv, Israel, Chicago’s O’Hare, Minneapolis-St. Paul, Dallas-Fort Worth, Phoenix Sky Harbor, and Detroit Metro are among those employing full-scale or trial systems with ALPR.
While toll roads, red-light violation and speed violation systems require a very high accuracy level for systems designed to automatically send citations to those who violate, parking is a different game entirely. The real test for effectiveness using ALPR in a parking application requires the plate to be accurately read on at least two different occasions: entering and exiting. (If you use equipment that allows you to automatically carry out the nightly License Plate Inventory, you may even have another accuracy read requirement.)
Given the requirement for multiple license plate reads, this may seem a more difficult task. But there are some things to consider that suggest it may, in fact, make the task of accurate plate recognition somewhat easier. Once the vehicle license plate has been read at the entrance gate, it becomes a captive number and one of a maximum fixed quantity. Let’s say we are dealing with a parking facility with 3,000 spaces. If a vehicle enters with a license plate number ABC123, it now has a potential of being mismatched with only those 2,999 other vehicles that may be in the parking facility. This could, of course, be more or less, depending on how many vehicles actually entered and exited during the duration, and assuming the lot was full for the whole time.
In practice, along with the license plate number, transaction number, date/time, and any other information stored each time the plate is read (such as the confidence level of each read performed), an image of the plate itself is kept on file. In this way, any disputes over the plate entering and the plate exiting can be resolved via some software that shows both plates on screen at the same time. This is referred to as “exception handling” — the only time the two images need to be seen is when there is an exception such as a mismatch at exit. This screen can be in the booth with an operator, or it can be at a central site, monitored by a supervisor as might be the case with unmanned exit booths where credit or debit cards are taken as payment.
An additional benefit of license plate information capture is one of security. If there is a particular license plate of interest to local law enforcement personnel, the parking facility’s data base can be queried to see if the plate is, or has been, on the premises. If there is a match, even down to just three or four consecutive characters, an image of the plate will be called up to allow the operator to verify that it is indeed the plate of interest right down to the state of issue, which can be determined from the image itself.
There now appears to be a good deal of interest in some less-than-traditional parking systems such as central-pay and pay-on-foot, and which may be under consideration for some larger parking facilities in North America. The advantages for using license plate capture technology becomes even more apparent in such applications. Parking fees in this type of system are paid by a person standing in front of a pay-station machine into which they feed their ticket plus some form of payment. There is no opportunity to read the license plate of the vehicle, which is parked at some distance from the pay-station. There has to be a mechanism to check the now validated ticket against the vehicle that is attempting to exit with that validated ticket. That ticket must be associated with the vehicle’s license plate against which it was issued upon entry. Ideally, it would be prevented from leaving — although the ticket presented has already been paid — at the exit gate if there is an exception such as a mismatch on the license plates. If it could not be prevented from exiting, perhaps it could indeed be treated as a “violator,” much as it would be in open-road tolling, and a citation could be sent.
Current indications would suggest that even checking license plates at exit against those taken from the manual License Plate Inventory (LPI), usually done in the airport operation off-hours, has great benefits in battling lost-ticket fraud. In many cases, the license plates are also manually put into the system at the vehicle entry point.
In addition to the lost revenue that is now recoverable because there is a license plate number in the system, imagine the additional savings in time and data accuracy that an automated system would provide. It would seem the wait-and-see period is over for this technology and ALPR is now an accepted part of the revenue management stream at larger parking facilities.
(Next month, we will discuss the issues of mismatching license plates.)
Jim Kennedy is President of INEX Technologies. Contact him through the company’s Web site (www.inextek.com).