License Plate & VIN Recognition

United Kingdom

Vehicle license number-plate, is a very important piece of information. It can give you, most of the upper layer details about the vehicle and its owner. 

Whereas a VIN number is a 17-digit number, which is a unique number to every vehicle. VIN gives the details of where the automobile was manufactured, details of the producer, details of vehicle pictures, type of motor, type of fuel used, potential of the motor, details of the security code which is given by the producer, details of the car manufacturing year and other car history details. There are several places where you can find your car’s VIN. The VIN positions are different for different car manufacturers and for car models.

Client

Automotive InsurTech Company

 

CLIENT

The United Kingdom is one of the largest automotive innovation hubs and home to few of the emerging AI adaptive insurTech startups. Our client is one of them, with a great vision to unlock the future of insurance with innovative technologies, solving auto insurance fraud detection and claims automation, while empowering insurers to deliver amazing customer experiences.

 

CHALLENGE

THE TOP 4 CHALLENGES:

  • Client didn't had enough data to give us for initial processing. So the only option left to us was to find and download publicly available data from internet. Finding VIN data was a very difficult task, since it is a very confidential information it was challenging to find the images of the VIN. But we managed to get some amount of VIN image data through various search engines like, Google search engine, Bing search engine, Yahoo search engine, Baidu search engine etc.  

  • Identifying different types of number plates like different color number plates, the number plate font, sticker kind of number plates was an another level of challenge. 

  • Same goes for VIN images as well, identifying different kinds of VIN numbers like the font, engraved digits, and embossed digits was quite a challenge. But using our Custom OCR, we were able to overcome this issue. 

  • After identification, here comes the part of OCR (Optical Character Recognition), identifying individual characters with blurred kind of images was a big challenge here. To overcome this, we have used image preprocessing techniques to solve this issue, which enhanced the image quality.

 

SOLUTION

We have used a custom Computer Vision Object Detection technique to identify number plates from vehicle images. After detecting the number plate region, we have cropped that number plate region and used computer vision techniques to enhance the cropped image and then we passed this image to the Custom OCR (Optical Character Recognition) model that we have built to identify the characters in the cropped number plate image. 

For the identification of VIN, we preprocessed the VIN image by enhancing it using computer vision techniques then we passed this image to the Custom OCR (Optical Character Recognition) model, the output contained all the text information from the VIN image, but we need only the VIN number, so to get that we used basic NLP and Regular Expression Techniques for extracting only the VIN number from the output text.

vin_and_license_plate_recognition_edited.jpg

Data Description: 

Number plate images and VIN images are used for this use-case. The vehicle images for the number plate detection model were collected from sources like Kaggle, and other open sources. The VIN images were collected from Google images and various search engines such as Google, Bing, Yahoo etc.

 

RESULT

We tested both the models on several images of number plate as well as VIN, We were able to identify all the different types of VIN, and Number Plates. The accuracy of our custom Number Plate Detection Model is around 98% and the Accuracy of our Custom OCR model on Number Plate is 92% and on VIN is around 93%.

Ready to put AI to work for your business?

Make a plan and understand your ROI before you start implementing AI. 
Don’t fall into the trap most companies fall into. 
Take the first step—Get in touch today.