Label Processing Engine for Automated Text Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text recognition technologies are inefficient in accurately reviewing product labels, especially those with non-Latin scripts and images, leading to errors and increased costs due to manual review requirements, and existing OCR technologies perform poorly with combined text and images, special characters, and languages written from right to left.
Innovation Solution
A system utilizing a label processing engine with object detection and text recognition modules, employing machine learning and deep learning models to detect and extract text from product labels, convert data formats, and compare extracted text with baseline data for accuracy, capable of handling multiple scripts and languages, including Latin and non-Latin scripts, and script written from right to left.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual review of product labels is performed, then accuracy can be maintained, but time consumption and human error increase
Solution Approach 1:
The patent replaces the manual mechanical review process with an automated optical character recognition (OCR) system that uses machine learning models to detect, extract, and verify text from product labels. The system substitutes human reviewers with computer vision technology that processes labels through multiple verification stages including object detection, text recognition, and baseline comparison, thereby eliminating human error while significantly reducing review time.
2Productivity
If existing OCR technology is used for label review, then automation is achieved, but accuracy deteriorates with non-Latin scripts and complex layouts
Solution Approach 1:
The patent segments the label review process into distinct stages: object detection to identify label regions, text recognition to extract characters, and verification to compare against baseline data. This segmentation allows each component to be optimized independently, with the verification stage specifically addressing accuracy issues by cross-checking recognized text against known baseline information, thereby improving overall precision while maintaining automation.
Solution Approach 2:
The system implements feedback through a verification mechanism that compares OCR-recognized text against baseline label data. When discrepancies are detected, the system can flag them for review or automatically correct errors, creating a closed-loop feedback system that continuously improves accuracy. This feedback loop enables the automated system to maintain high precision even with challenging non-Latin scripts and complex layouts.
Data Source
AI summary
A label processing engine receives, as inputs, raw data representative of a label and baseline data, the product label being unique to at least one specific product and the baseline data corresponding to the at least one specific product. The engine detects a raw data object within the raw data, classifies the raw data object, and localizes the raw data object within the raw data. The engine detects a baseline data object within the baseline data, classifies the baseline data object, and localizes the baseline data object within the baseline data. The engine recognizes corresponding text within the raw and baseline data objects and extracts the corresponding text, reassembles the corresponding text into respective lines of text, compares the respective lines of text with one another, and generates a tabular and/or visual report including a notification based on the comparison.