Upload-Time Document Type Detection Using Dual OCR Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing document upload processes in online computer systems fail to accurately and efficiently identify document types during the upload process, particularly in backend systems that require specific document types for processing, leading to inefficiencies and potential fraudulent activities.
Innovation Solution
A two-step optical character recognition (OCR) process is implemented, utilizing a keyword recognition engine and an account character recognition engine to identify and verify document types, with the backend system initiating detection during the upload process, allowing for real-time validation and cancellation of unsupported document types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a backend system processes millions of documents, then processing volume increases, but speed and accuracy in determining document types become critical challenges
Solution Approach 1:
The system performs preliminary actions by detecting document types during the upload process itself, before the actual processing begins. The backend system initiates character recognition engines to identify document types in real-time as documents are being uploaded, rather than waiting for the upload to complete and then determining document type. This preliminary detection during upload significantly reduces the time loss associated with separate document type determination steps.
2Measurement precision
If multiple character recognition engines are used to detect and verify document types, then accuracy improves, but system complexity increases
Solution Approach 1:
The system segments the document type detection function into multiple specialized character recognition engines. Instead of using a single complex recognition system, the patent divides the task into separate engines (e.g., one for detecting keywords like 'check' or 'deposit slip', another for verifying with account characters). Each engine handles a specific aspect of document type identification, and their results are combined through verification logic. This segmentation improves overall accuracy while keeping each individual engine relatively simple.
Solution Approach 2:
The system implements feedback through a verification process where the results from multiple character recognition engines are cross-checked. The second character recognition engine verifies the document type identification by detecting account characters and comparing them against the initial detection. This feedback loop ensures accuracy by validating the results from the first detection process, allowing the system to correct potential misidentifications.
3Reliability
If real-time document type verification is performed during upload, then fraudulent activities are reduced, but processing speed may be impacted
Solution Approach 1:
The system performs preliminary document type detection during the upload process itself, before processing begins. By initiating character recognition engines when the upload starts and obtaining results during the upload transmission, the system can reject fraudulent documents before they enter the processing pipeline. This preliminary action maintains security without requiring separate verification steps after upload completion.
Solution Approach 2:
The system maintains continuity of useful action by performing document type detection concurrently with the upload process rather than sequentially. The character recognition engines operate in parallel with the file transmission, and the verification process continues as the upload completes. This continuous detection ensures that fraudulent documents are identified and rejected without creating a separate processing bottleneck, thereby maintaining upload processing speed while ensuring reliability.
Data Source
AI summary
Systems for document type detection during a document upload process are disclosed. Character recognition processes are performed on images of a document being uploaded to a backend system. The character recognition processes may identify different subsets of characters in the document which are used to identify the document type. Subsets of characters may include keywords and account characters and each subset may be processed by a different character recognition process. One or both of the subsets of characters may be used to identify the document type and the identified document type may be used to decide whether to proceed with the document upload process.