Receipt Estimation via Barcode Cross-Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating purchased items from receipts, such as OCR-based character recognition, face challenges due to character recognition errors, limited printed regions, and unique retailer notation, leading to incorrect database matching.
Innovation Solution
An information processing system that combines receipt information from a captured image with code information from one-dimensional or two-dimensional codes, allowing for accurate estimation of purchased items by matching character strings against database entries associated with detected codes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If OCR-based character recognition is used to read receipt information, then the process can be automated, but recognition errors occur due to limited printed regions and unique retailer notation
Solution Approach 1:
The patent introduces code information (barcodes, QR codes, or character strings) as an intermediary element that bridges the receipt and the database. Instead of directly relying on OCR to match article names, the system uses detected code information to retrieve corresponding article names from the database, which then serve as reference for correcting or verifying OCR results. This intermediary approach resolves the contradiction by maintaining automation while improving recognition accuracy through cross-validation.
2Productivity
If database checking is performed using OCR read character strings, then article estimation can be achieved, but matching fails due to recognition misses and retailer-specific notation variations
Solution Approach 1:
The system implements a feedback mechanism where code information detection results are used to verify and correct OCR recognition results. The detected code information retrieves article names from the database, which are then compared with OCR results. When discrepancies are found, the system can either correct the OCR results or select the more reliable source, thereby improving database matching reliability while maintaining processing efficiency.
Solution Approach 2:
The patent performs preliminary detection of code information (barcodes, QR codes, or character strings) on the receipt before attempting database matching. This preliminary action retrieves the correct article names from the database in advance, creating a reference set that guides subsequent OCR result verification. By preparing this reference information beforehand, the system ensures reliable matching while maintaining efficient processing.
3Device complexity
If only receipt character strings are used for article estimation, then the process is simple, but accuracy decreases due to recognition errors and notation variations
Solution Approach 1:
The patent merges multiple information sources - OCR read character strings from the receipt and code information (barcodes, QR codes, or character strings) detected from the same receipt - into a unified article estimation process. By combining these sources and cross-referencing them with database entries, the system achieves accurate article estimation without significantly increasing system complexity, as the additional detection and matching steps follow a straightforward workflow.
Data Source
AI summary
Provided is an information processing apparatus including an estimating unit to estimate a purchased article indicated by receipt information on the basis of the receipt information indicating a result of reading characters from a first captured image which is obtained by capturing an image of a receipt, and code information indicating a result of detecting a one-dimensional code or a two-dimensional code from a second captured image which is obtained by capturing an image of a one-dimensional code or a two-dimensional code of an article.