Data on vehicle registration certificates may be needed for document series registration in business and government structures. But to bring them into the system, you have to fill in too many margins manually: numbers and letters of the code, the number of the registration certificate and other data. It takes much of time, especially for clients who issue an electronic policy by themselves.
A common mistake can be made in each margin by the manager or the client himself and bring difficulties for further work.
The Goal of the Documents Scanning System (Recognition)
Make the design process as simple, convenient, fast as possible, and minimize the possibility of error. Ideally, the document registration should take only a few seconds to complete and to enter the data enough to take a photo of the vehicle registration certificate on both sides.
A Few Words about the Development of Registration Certificates Recognition Technology
We used unique Evergreen designs in the field of image processing, algorithms for obtaining a response from libraries using an API key, as well as publicly available technologies to create a document recognition system.
Since each document has different margins values and their quantity, there are no universal solutions. We used algorithms based on neural networks to achieve the best result.How we taught registration certificates recognition system correctly recognize the disputed data
When creating the system, we took into account that the human and computer eyes work differently. For example, a computer doesn’t perceive the prospect and the letter "I" at a particular inclination can be seen as a point or a number "1". Moreover, since the machine can’t realize the error, it enters the data as it saw.
To avoid such mistakes, we taught the system to recognize margins using the advantages of artificial intelligence. For example, consider that in the first and last name may not contain numbers. After appropriate training, the system learned to recognize margins where only numbers or letters are possible. Therefore, if recognition system sees in the name or surname margin the number "1", it replaces it with "I", and the number "0", with the letter "o". Also, depending on the type of margins, the system recognizes with what kind of registration certificate it deals with: a new or old design.
Photo verification by the registration certificate recognition system:
1. Image acquisition
2. Analysis of its position in space
3. Image Adaptation for Recognition
4. Finding and processing margins
5. Recognition of information from the document margins
6. Processing of recognition results (verification)
7. Preparation of the answer
8. API response and its work with CRM client directories
What it looks like for customers:
Document Scanning Can Be Integrated with Business Systems
Since each business works with its customized system (CRM, 1C, etc.), integration always takes place individually. Also, we take into account that often the clients reference books are already configured in the clients systems.
The integration of the registration certificates recognition service into the client's CRM system may look like in such a way that when you fill out the required document, the “load image” button appears. Thus, both customers and service employees get an intuitive understanding of how to use the updated system.
It is also possible to integrate the system with chat bots, which will allow the staff to save time and automate the document processing. A chat bot can communicate with a client through your website as well as through a Viber, Telegram, Skype, and other communication channels and recognition system will make registration fast and straightforward.
Photo Uploading Rules for Accurate Registration Certificate Recognition
You can upload a photo of a registration certificate into the system in two ways:
take a photo from the phone, tablet, if the order is processed online from a mobile
make a photo from the webcam, if the order is made from a laptop or a desktop computer in the insurance office.
So, in the Photo:
- all 4 corners of the document should be visible;
- no noticeable highlights should be on the margins that are recognized;
- it is advisable not to mess up the viewpoint;
- the top of the passport should be perpendicular to the horizon line;
- uniform background and lighting;
- the registration certificate should occupy from 40% to 70% of the total photo size.
With all these recommendations, the recognition accuracy of the datasheet is 92-94%. Considering the likelihood of users deviating from the standard, we have created algorithms that allow the system to turn the image if it is upside down or at an angle, which improves recognition accuracy.
Image processing after loading lasts from 2 to 20 seconds, depending on the quality of the photo. At the end of processing, the user sees a screen with the margins of his profile already filled in. As soon as the correctness of the recognized data is confirmed (and, if necessary, minimal corrections are made), it remains only to print out the contract and sign it.
Who Can Use the Program for Scanning Registration Certificates
The speed of processing the necessary documents is increased several times while using registration certificate recognition system. The convenience for the manager and the client is increased, the factor of human errors is reduced.
If you need an individual solution, we gladly inform you that our system of recognition of registration certificates has artificial intelligence embedded in it, respectively, we can easily modify it for your needs. Interested? Feel free to contact us and get a system that will increase the efficiency of your business exponentially.