In this article, we elaborate on the need for ANPR in India and the unique challenges to high-accuracy ANPR for Indian license plates. We also share some specific examples of how we have tuned our ANPR algorithms to support the beautiful country of India.
What is ANPR?
Automatic Number Plate Recognition (ANPR), also known as Automatic License Plate Recognition (ALPR), is software used to recognize the number plates automatically by performing sophisticated optical character recognition on images to read the license plates of vehicles.
A typical ANPR solution takes an image as input and returns the decoded vehicle plate, timestamp and confidence level of the decoding algorithm. The image is analyzed in as fast as 50 to 100 milliseconds to enable real-time surveillance.
Need for ANPR in India
Unlike other countries, India, with its one billion people population, has a unique set of needs for ANPR. The main use of ANPR is in highway monitoring, parking management, and neighborhood law enforcement security.
In India there is one death in every four minutes with most of them occurring due to over speeding. ANPR is used to monitor the vehicles’ average speed and can identify the vehicles that exceed the speed limit. In this case, a fine ticket can be automatically generated by calculating the distance between two cameras. This helps to maintain law and order which, in turn, can minimize the number of road casualties.
ANPR provides the best solution for providing parking management. Vehicles with registered plates can automatically enter into parking areas while non-registered vehicles will be charged by time of check in and check out. Number plates of the car can be directly linked with owner mobile phone and parking tickets could be paid without any extra effort directly from the user’s account against the ticket number generated. ANPR can support a cloud-based system pre-book and pre-pay platform for parking.
In India 200,000 cars are stolen per year. This number can lessen if proper steps are taken and ANPR system is used to track cars so that if vehicles are stolen, law enforcement will be able to identify when, where and the route taken by a stolen vehicle. This can help bring justice swiftly to such a vast nation.
India license plates are as diverse as the country itself–lots of styles, colors, fonts, sizes, etc. Source: Plate Recognizer Clients.
ANPR Challenges in India
The need for ANPR in India is strong. However, the solution may not be as clear.
ANPR in countries such as Vietnam, Australia and Italy, which have fairly standardized license plates, have accuracy levels often exceeding 90%.
India, however, is quite different. The variations of makes it difficult for an ANPR system to accurately identify and decode the license plate. In India, there are 210 million vehicles with over 50 different license plate types. The plates vary in style, color, fonts, sizes and even location in the vehicle. For example, certain regions of India would have certain plate formats and certain generations of vehicles have their unique peculiarities.
Moreover, cameras deployed in India tend to be of lower quality, thereby compounding the ability for ANPR engines to accurately decode the license plate. The low-cost cameras have limited visual coverage, are equipped with less acute motion and object detection sensors, and contain limited night vision capabilities. These three factors increase the number of blind spots and occurrence of blurry images. This situation is made worse during nighttime and inclement weather when visibility is lower.
As such, ANPR for India is particularly difficult and accuracy rates seldom exceed 70% from our internal analysis of our and competitive ANPR providers.
Plate Recognizer Improvements for India ANPR
We at Plate Recognizer appreciate the importance of ANPR in India. After building our “global” ANPR solution, we focused our efforts on improving plate recognition in India, since we received quite a lot of interest from developers in that country. Little did we know, though, about the all challenges as mentioned above for the India license plate recognition market.
Here is our 5 step process to improve India ANPR:
1) Define License Plate Variations. We began our journey by spending a lot of time assessing all the license plate variations across the different regions of India. We learned that the plates were as diverse as the people groups, languages and cultures within the country!
2) Create Regex. For each license plate variation, we created the plate’s regex (tech term for regular expression). We also analyzed and defined the font type and font size of each license plate type.
3) Get Sample Images of India License Plates. Thanks to our partners in India, we received and manually processed thousands upon thousands of images of real-life India license plates and fed that into our ANPR engine. The images contained mopeds, motorcycles, aged vehicles, new sedans, station wagons, minivans, light-weight trucks, semi-trailer trucks, flatbed trucks, buses and more.
4) Analyze License Plate Patterns. Our plate recognition engine matched the regex to each plate type and image. This allowed us to “educate” the engine with specific parameters that pertain to that vehicle and plate type.
5) Assess Improvements. Once a pattern has been established, we reran the base images through the ANPR engine to see what improvements were realized.
Over the course of several months, we iterated each day and found small, but encouraging improvements to the ANPR accuracy rates. Little by little, plate by plate, we inched up the ladder.
As we saw the improvements, we applied our global set of algorithms to compensate for the photo quality and lighting conditions. This allowed our number plate recognition engine to improve its accuracy for low resolution, blurry and dark images.
After months of intense work, we were able to take our ANPR accuracy rates to upwards of 90%. Depending on the environment and situation (e.g. parking lot in city center), our customers are seeing over 95% ANPR accuracy for India.
After applying our 5 Step Process to Improve ANPR for India, we were able to successfully recognize these and other plates. Source: Plate Recognizer.
Other Ways to Improve ANPR
In addition to enhancing the underlying ANPR engine for India, we have made some concrete suggestions to help our customers achieve high accuracy ANPR with Plate Recognizer.
1) Adjust Motion Detection. Depending on the camera and/or Video Management Software (VMS) used, we see images where the vehicle is barely visible. This may be due to the fact that the motion detection is too sensitive or that the camera and/or VMS is capturing the image right when motion is detected. Fixing this issue is easy: just set a slight delay in the image capture. So rather than immediately capturing the image, wait 0.5 seconds for the vehicle to be more prominent. This way, the number plate can be more present in the photo for the ANPR engine.
2) Focus Camera. It is important to focus the ANPR camera on the incoming vehicle plate. We often see cameras in front of parking lots that aim wide across the parking lot rather than towards the entrance or exit areas. Again, the fix is simple: just adjust the camera position and width of focus so that the license plate can be more present.
Rather than focusing camera on entire region, ANPR can be improved by focusing camera towards the red box highlighted above. Source: Plate Recognizer Client.
3) Proper Lighting. We have made great inroads with our ANPR engine to decode license plates in dark environments. However, in some situations, clients share images with us where the headlights make it impossible to even detect that a plate is present. To fix this, you’ll need to brighten the entrance/exit points or use a camera with higher aperture.
While the suggestions above pertain to all countries, not just India, they can significantly improve accuracy rates for ANPR if implemented correctly.
Huge Opportunities for ANPR in India
As India goes through a phase of modernization and infrastructure development in the upcoming decade, we see significant opportunities for the various government entities, municipalities, and businesses to embrace ANPR.
In the meanwhile, we will continue to invest in our ANPR engine specifically tuned for India so that we can make it better and better!
You Can Help!
Are you working on an ANPR project for India? If so, you can help!
With more vehicle images, we can continue to tune the engine to address your use-case for ANPR. Just send us 5,000 to 10,000 vehicle images via Dropbox or FTP. Together, we can make ANPR for India stronger and stronger. for more info!