Using computer vision to verify identities
Background
Analytics Engines are supporting are supporting an organisation that provide an identity verification services to public and private sector organisations. Their platform provides background verification solutions for businesses. In 2021, the client was awarded a contract to provide digital ID checks as part of NI Direct’s Covid Passport process.
As part the NI Direct COVID Passport Process, the client was required to verify the identity of individuals using various forms of identification including passports, electoral cards, and bus passes.
The client noted that government departments and organisations used a wide range of identity services, resulting in people having to enter the same information multiple times to access multiple services. This led to mounting inefficiencies, poor user experience, lack of business control, and reduced capability to tackle fraudulent activity.
Processing these failed identity checks required manual human intervention; a process that is expensive and time intensive. The client required a solution that could perform these identification checks effectively and efficiently.
The solution
Analytics Engines have created an AI-based document classification module that can classify between various forms of identification including passports, bus passes, and electoral cards. The solution is considered a ‘multi-class’ classification module.
At present, the solution has been designed to distinguishing between 8 (with the ability to add more) separate identification documents. In addition, the solution utilises these same technologies to identify forgery and potential manipulation within submitted documents.
The solution is able to automatically process documents with greater effectiveness and efficiency than a human counterpart, reducing expenditure on manual intervention from 3rd party suppliers.
The solution can identify and verify the uploaded document type in milliseconds with a high degree of accuracy. The solution has also been designed so that it can be easily integrated with the client’s existing products and platforms.
Standard classification techniques don’t take into account out of set examples. Out of set entities are entities not accounted for in the initial classification. We introduced additional qualification to ensure that any upload documents aligned with expected documents.
To ensure the consistency and quality of the images received (processing and standardised), we also implemented a cropping technique to eliminate variability such as background in images.
