Neural Graph Receipt Decoding for Wrinkled Paper Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional optical character recognition (OCR) systems struggle with decoding receipts that are wrinkled or deformed, and those with large gaps between data items, due to misalignment caused by image and paper distortions.
Innovation Solution
The proposed solution involves a line detection framework that uses a neural graph architecture. This framework includes vertex feature representation circuitry, graph neural network circuitry, and post-processing circuitry to generate an adjacency matrix and correct text alignment, effectively handling misalignment and gaps in receipts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional OCR systems are used to decode receipts, then the decoding process is simple and fast, but the text recognition accuracy deteriorates on wrinkled or deformed receipts with large gaps between data items
Solution Approach 1:
The patent segments the receipt processing into distinct stages: initial OCR text extraction, vertex feature representation, graph neural network processing, and post-processing. This segmentation allows each component to specialize in handling specific aspects of the problem, improving overall accuracy while maintaining manageable system complexity through modular architecture.
Solution Approach 2:
The patent introduces an intermediary graph neural network component that bridges the initial OCR extraction and final text recognition. This intermediary processes the extracted text through vertex feature representation and adjacency matrix generation, enabling the system to handle distorted receipts with gaps by establishing contextual relationships between text elements.
2Adaptability or versatility
If traditional OCR systems are used, then the system complexity is low, but the ability to handle misalignment and gaps in distorted receipts deteriorates
Solution Approach 1:
The patent transitions from traditional linear text processing to a graph-based dimensional structure. By representing text elements as vertices and their relationships as edges in a graph, the system gains the ability to capture spatial and contextual relationships in multiple dimensions, enabling it to handle misalignment and gaps in distorted receipts effectively.
Solution Approach 2:
The patent changes key parameters in the text processing pipeline by introducing polar coordinate calculations for vertex positioning, adjacency matrix generation for relationship modeling, and clique-based grouping for text line reconstruction. These parameter changes enable the system to adapt to various distortion types while maintaining structured processing.
Data Source
AI summary
Methods, apparatus, systems, and articles of manufacture are disclosed to decode receipts based on neural graph architecture. An example apparatus for decoding receipts includes, vertex feature representation circuitry to extract features from optical-character-recognition (OCR) words, polar coordinate circuitry to calculate polar coordinates of the OCR words based on respective ones of the extracted features, graph neural network circuitry to generate an adjacency matrix based on the extracted features, post-processing circuitry to traverse the adjacency matrix to generate cliques of OCR processed words, and output circuitry to generate lines of text based on the cliques of OCR processed words.