ANPR, also known as Automatic Number Plate Recognition, (or Automatic License Plate Recognition, ALPR, in some countries), is software that decodes vehicle license plates. While it previously operated purely on stationary cameras (think roadside cameras or closed-circuit television), this type of technology has started to evolve.
In this article, we will share some insights and advantages of our approach to deploying ANPR solution on a mobile phone.
The Need for ANPR on Mobile Phones
Whether you’re interested in parking management, vehicle inspection or fleet management, ANPR technology is an important ingredient in your solution. ANPR for mobile phones puts the technology directly in the hands of your users and enables the technology to be used in a variety of environments.
For example, parking patrol personnel can use ANPR on their mobile phones to identify and tag vehicle citations for expired parking. Vehicle service personnel can use their Android phones or iPhones to take images of vehicles involved in an accident and also capture the vehicle license plate to appropriately identify the car.
Common ANPR Solutions on Mobile Phones
The current ANPR technology on mobile phones consists of mainly of basic Optical Character Recognition (OCR) technology. OCR technology essentially converts any type of scanned file or image into editable or searchable data. For example, if you were to use an app on your phone to scan a PDF document, OCR technology would allow you to search the document for information once it was uploaded.
However, when it comes to ANPR, it can become difficult to use OCR technology. OCR has a basic set of algorithms that operate best with particular clear images. This requires images that are sharp, have solid contrast, accurate coloring, and more. All of these elements can often be difficult to achieve for the variety of license plates that exist in your region–let alone factors such as motion, weather or other obstacles. The result is a really low accuracy on your ANPR efforts via an Android phone.
The next section details the steps to take if you’d like to utilize ANPR correctly on a mobile platform.
With Plate Recognizer, great, high-accuracy ANPR is possible on a mobile phone! Source: Plate Recognizer.
Best Way to Get ANPR on Mobile Phone
One of the greatest benefits of a mobile phone is that your phone is connected to the Internet. This means you can use the power of the cloud to get high-accuracy ANPR rather than rely on an inferior ANPR (or rather OCR) engine on the phone.
The steps to get ANPR on your mobile phone are simple as 1, 2, 3:
- Take Image of License Plate. Every smartphone has a camera. So use it! You can either use your phone’s native camera or wrap a camera within your own mobile app.
- Send Image to Plate Recognizer. Once you have a photo of the vehicle with the license plate, you can send that photo straight to Plate Recognizer’s ANPR SDK or API Cloud.
- Get Decoded Plate. In as fast as 50 milliseconds, you’ll get the decoded license plate!
That’s it! Just three steps to get high-accuracy ANPR via a mobile phone.
How Plate Recognizer is Made for ANPR on Mobile
We have created our automatic license plate recognition service to uniquely serve users on mobile phones. Here are some things we’ve done:
- Blurry, Dark Images. It’s not always steady hands that hold a mobile phone. We have optimized our ANPR engine to decode license plates on blurry and dark images.
- Confidence Level. We specifically constructed a Confidence Level for each and every license plate image. This way, you have a heads up on whether or not to let your users resubmit the image.
- Fast. In the age of instant access and on-demand services, our license plate recognition engine decodes the plate in near-lightning speed, that is 50-100 ms.
Tips to Improve ANPR on Mobile
We have a few suggestions on how to use a mobile phone for license plate recognition:
- Decrease Image Size. It is also recommended that you decrease the image resolution size to minimize the data required to send the image to the ANPR engine. To get an accurate license plate reading, the license plate only needs to be 40 pixels in width. So if the user is up close to the vehicle, then a photo resolution of 640×480 would be more than sufficient. By decreasing the image size, you don’t have to send large files over your wifi and/or the Internet.
- Use SDK for Speed. If you need to decode the license plate image very quickly, then go with the Plate Recognizer On-Premise SDK. The SDK can return the license plate in 50-100 milliseconds. This way, you’re sending the images within your environment and not over the Internet.
- Use API Cloud for Simplicity. It is super-easy to get the Plate Recognizer API Cloud up and running with your mobile phone, since all the ANPR source code is already provided.
- Minimize Storage. No need to store the image in the phone’s drive. The image data can be an in-memory variable, so you do not have to write the images to disk before calling our API.
- Allow Users to Retake Photo. Sometimes, the decoded plate is not accurate because the lighting of the photo may be poor, a leaf has obstructed one of the license plate characters, there was an error in the transmission of the image, etc. The best thing about Plate Recognizer is that its license plate recognition algorithms return the Confidence Level of the decoded license plate. So if the Confidence Level returned is under 80%, you can inform the user to take another photo of the license plate. Thankfully, if the license plate is properly photographed, your users would have to do a re-take in less than 1-2% of the time.
- Landscape Images. Our ANPR engine is more optimized for landscape images. So rather than having users take images in portrait, you’ll see a faster, more accurate response in the decoded license plate with a landscape image. After all, a license plate is definitely more landscape than it is portrait!
Benefits of Our Approach to ANPR on Mobile Phones
The benefits of this approach compared to other ways of deploying ANPR on your Android or iPhone devices are many:
- Fast to Market. Writing the software module to grab and send the image from the phone camera to Plate Recognizer can be done in a couple of hours. You can use the code we already have in Shell, Ruby, Python, Javascript, C#, etc.
- Super Low Cost. Because all ANPR mobile requires is a phone camera, this approach can work with basically any mobile phone that has access to the Internet. There’s no need to upgrade your mobile device or be concerned about the phone’s memory or processing power. Use the phone you already have!
- Best Accuracy Rates. The ANPR engine itself is already very accurate in decoding the license plate. But with the added insurance of the Confidence Level (see Tip #4 above), you should be able to achieve 100% accuracy on your ANPR.
- Quick and Efficient. The whole process (from taking the photo to the user-confirmed decoded license plate) takes less than a second. You’ll have everything you need in just a couple of clicks.
- Integrated GPS. Need to know exactly where the vehicle is located? No problem. Just make sure to pass along the GPS coordinates of the image in your call request to Plate Recognizer API Cloud or SDK.
Parting Thoughts
Great, high-accuracy ANPR is possible on a mobile phone! With our approach, you can achieve this quickly, easily and cost-effectively on your Android or iPhone device. Leverage your phone’s connectivity and get decoded license plates in seconds with Plate Recognizer Cloud API and On-Premise SDK. This also works for other countries like our ANPR for Brazil.