4 minute read
Back to the future
Innovative license plate recognition technology and a customer focus are proving to be a successful combination
Words | Mate Kiss-Gyorgy, Asura Technologies, Hungary
We live in an era where the customer is truly king; not even niche technology providers or manufacturers have the luxury to leave their customers to sink or swim after purchasing their products. Competition is fierce and companies who fail to listen to their audience are quick to fall. At least this is how it should be, but in truth, there are industries where this shift seems still a long way away – and the license plate recognition (LPR) industry is one of them.
Why, you ask? The reason is simple: most technology providers do not regard integrators as customers. Rather they are considered developers, who have the means, personnel, time and resources to develop their LPR application from scratch. Which could not be further from the truth.
Change is on the horizon
Up until now, the customer-oriented IT evolution of the past decade has avoided the LPR industry, but that is about to change.
Typically, an integrator purchases the optical character recognition (OCR) engine and from that point on, they are left alone. This is the reason why integrators need creativity, time and lots of patience to learn to use these products. Support is limited and no warranties are included – application programming interface (API) sample codes are provided but, unfortunately, in most cases, they will not get you a working LPR system.
The advantage of such a setup is flexibility in terms of usage, yet the integrator needs to do lots of system development. Why? An engine module is usually capable of processing one image at a time and returns the result in the memory (RAM) or as a database entry.
Several critical functions also need to be added. These include parsing the stream of the image source (IP camera), filtering irrelevant images to obtain reasonable and affordable resources, loading and processing images, implementing the thread handling and license optimization, managing and validating events, and finally implementing a database solution.
A long time in the making
Developing an application with all these features takes years, even for OCR software providers. It takes several years of work of 10 full-time developers to deliver just such a high-efficiency LPR solution
Furthermore, the licensing model of such a solution will not match the integrator’s business model. Thread/CPU core-based licensing is unpredictable: too many variations for processing time and too many images handled per event. To make it failsafe, the technology provider will set up a costlier than necessary system, taking the worst-case scenario as the basis of their calculation.
An increasingly popular way to integrate LPR is to purchase cameras mounted with LPR software. Integrators choosing this option usually face performance and cost issues.
Saving setup costs and system design
With smart design, a high-end CPU computer is capable of processing the footage of 16 cameras, in some cases even more. If image
processing is performed using an IP camera, you have to say goodbye to that processing power. You will only get a fraction of a computer’s capacity, meaning that from a certain project scale you will need to buy more cameras. This brings us straight to cost. When onboard units are responsible for image processing, an OCR license needs to be purchased for each camera – which is a waste considering that in a setup where a computer processes the images, only one or two licenses may be required for 16 cameras.
So why not leave out the server as it is not needed for processing – if it is done by the camera? The answer is easy: you need a server anyway. The data cannot be stored on these cameras permanently, so you need a server for centralized data collection. This server can run the LPR application and perform the processing as well. Why would you unnecessarily duplicate the resources required for the operation?
Another downside of using such system design is that the maximum recognition rate you can achieve is 95%. In difficult regions or countries, this might drop down to 80% or less, even when using high-end, expensive products.
Customer oriented LPR solution
An LPR solution is generally expected to perform three tasks: connect to an image source with a live stream of the footage; maximize recognition rate with the least resources by optimizing performance; and provide validated results in database format that fit the integrators’ requirements.
Asura’s LPR does all this in one tidy package with recognition rates of 97-99%. It is cheap, easy to integrate, and has all the necessary functions embedded to be an attractive LPR system.
The increasing urbanization of many countries brings new challenges in fields relying heavily on LPR. These include tolling, ITS, surveillance, congestion charging, traffic violation enforcement, traffic monitoring and journey time measurement.
New service providers enter the market expressing an increasing demand for outsourcing the hosting and operation of LPR, paving the way for Software as a Service and other cloud-based, solutions. We are in the middle of a transition from product-based traffic to service-based solutions.
Asura Technologies believes that services must be designed to fit customer expectations. With a visionary mindset, things can be reinvented, (mis) conceptions can be changed and boundaries are only there to be expanded. Asura Technologies created LPR technology with integrators as its customers in mind. Its business model centers around the customer, who needs to be reassured by warranties, and must be provided the latest technology and high-end customer service. n
Left: Asura applies Software as
a Service (SaaS) standards to its license plate recognition technology
Bottom left: Even in cases of