No more Cheaty Accuracy Rates in ALPR

Innova certificates an accuracy rate of 99,3% without discarding any image during the analysis.

The accuracy rate is a key factor in parking management, especially when we are implementing barrierless or ticketless solutions. In the ALPR market, different technological solutions guarantee high rates when referring to accuracy and efficiency. How is this metric calculated? How do manufacturers measure accuracy rates in ALPR systems? Simple: LPR Manufacturers usually hold their statistics discarding the images of the damaged license plates and offering an accuracy rate only from the good plates. But at Innova we can assure an accuracy rate higher than 99% without discarding any image during the analysis… and we can guarantee in real installations. 

At Innova, we have been working from the beginning trying to solve the accuracy rate calculation issue and we have finally succeeded thanks to the deep learning technology from OCR5 and its capability to work with colour images using white light. With Innova´s technology, license plate defects will not be a problem of efficiency and profitability. Our solutions guarantee maximum readability of all types of license plates regardless of their condition, entry/exit angle or lighting. The use of colour and an ALPR solution based on Deep Learning algorithms improves the success rate without doubling the installation cost because, with a single camera, the success rate is maximized.

In order to obtain real information about our technology and accuracy rate, we held this test with S&B Iberica in an important German discount retailer chain that has more than 12.000 stores over the world. After the analysis of more than 4000 images over 3 days the result obtained was 99,3% of Accuracy Rate. Along with this evaluation, we did not discard any pictures due to any plate damage, which leads us to an accuracy rate as real as possible without gaps when evaluating parking’s real performance.

This accuracy rate of 99,3% breaks with the trend to work using IR light with a bandpass filter to work very focused at the IR spectrum. The reason why a lot of LPR manufacturers are anchored in this old-fashioned trend is that they are using old neural networks algorithms and the time expended to find the plate at the image is drastically reduced with a very high contrast image. But the problem with working very focused on the IR spectrum is that the character recognition algorithm is very sensitive to the damages at the reflective emulsion of the plates.

With the development of OCR5, the first LPR solution based on Deep Learning, we discovered that the most advanced deep learning algorithms spend the same time searching for a plate in a highly contrasted image as in an image in the visible spectrum where there are more objects to discard. This is the reason why we started in 2014 to work without any IR filter in the camera and in 2021 we launched the new Siram Compact Hybrid, which can work in the IR spectrum without any filter or in the visible spectrum using a white light projector and a colour camera.

Regarding the test held, thanks to S&B Iberica, we can guarantee an excellent result: a 99,3% accuracy rate without discarding any damaged licence plate image.  Are you willing to risk the profitability of your parking on cheaty rates? Are you looking for a technological solution to trust the correct functioning of ticketless and barrierless systems? Do you need a higher accuracy rate in your car parks? 

You are one click away from achieving maximum efficiency and profitability in managing your parking lot. Contact us!

SIRAM MAKE