Traditional KYC procedures are infamously labour-intensive. They generally involve making a trip to a branch office and manually filling out forms that are time-consuming, costly, and limited to office hours.
Customers today expect instant gratification and will accept nothing less than real-time digital services round the clock. With digital competitors lurking around every corner to entice dissatisfied customers, it is imperative that companies don’t appear to be creating unnecessary hurdles whilst adhering to regulatory guidelines. This, coupled with the physical challenges of conducting business-as-usual during a global lockdown, creates more room for innovation than we’ve ever witnessed before.
Innovation in Isolation
Optical Character Recognition (OCR) is a technology that recognizes text within a digital image and is used to convert virtually any kind of images containing text into machine-readable data. Prior to OCR technology, the only way to digitize printed paper documents was to re-type the text manually. This was not only time consuming but also increased the possibilities of human error.
While OCR is commonly used to recognize text in scanned documents, its technology is also used in data entry automation, indexing documents for search engines, automatic number plate recognition, as well as assisting the blind and visually impaired. It has been exceedingly useful in transforming historic newspapers and texts into searchable formats by digitizing them.
Optimizations through Machine Learning
A few common problems encountered by OCR over the years, such as blurs, glares, incorrect image capture are now being tackled by the use of Machine Learning. Obviously, conversion accuracy is important, and most OCR software delivers 98% to 99% accuracy.[1]
OCR solution by Kwik.ID:
The OCR solution offered by Kwik ID enables instant verification of OVD’s like Aadhaar, PAN, VoterID, Passport, and Driver’s License. It makes it easier to detect frauds through signature matching tools and leverages ML to spot problematic information. Furthermore, Kwik.ID also promises its customers of Video KYC at low bandwith with its collection of the Best Video KYC tools.
We have built a state of the art OCR engine that extracts meaningful information from the image of an identity card using the following module
- Rotation and cropping of Image
- Raw text detection from natural image
- Extraction of meaningful data from Raw text
- Rotation and cropping of Image
For the text detection module to work properly, it’s essential to remove the background noise and make the ID card image horizontal with 0 degrees. However, since the image of the ID card is provided by an end-user, it may contain a lot of background noise & can be at any angle. This module helps in the perfect alignment of the image & removal of the background noises.
2. Raw text detection from natural image
Once the card image is cropped and rotated, it is provided to the ML engine that detects text in the image. These detected texts are then extracted from the image along with their corresponding coordinates.
3. Extraction of meaningful data
In the final step, we extract meaningful information from the detected text in step two. However, just detecting raw text is not enough. We need to identify which text represents which information. For example, if two names are detected in an image, which one is the father’s name, and which one is the card holder’s name should be segregated.
Here the coordinates extracted with each text play the role. Based on different types of cards, the extraction engine contains different templates for each type of ID card to identify the position of details.
While not a new technology, OCR has been fine-tuned via AI & ML making it an indispensable tool for quicker customer onboarding and thereby ensuring smoother Video KYC verification.
To know more about Kwik.ID’s seamless e-KYC solution, please write to info@getkwikid.com.
[1] https://tdwi.org/articles/2018/03/05/diq-all-how-accurate-is-your-data.aspx