AI Invoice Decoding with Column Detection and Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional invoice recognition systems struggle to accurately decode invoices from non-scanned images, which are of lower quality due to inferior lighting, focus, framing, and resolution, and fail to extract tracking information such as item types, costs, and quantities.
Innovation Solution
The use of artificial intelligence, specifically natural language processing, computer vision, and deep learning with region-based convolutional neural networks to detect columns and classify headers in images of invoices, enabling the extraction of purchase facts and conversion of image data into digital format.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional invoice recognition systems are used to decode invoices from non-scanned images, then the system is simple and easy to implement, but the accuracy of decoding is poor due to inferior lighting, focus, framing, and resolution
Solution Approach 1:
The patent segments the invoice decoding process into multiple specialized AI models: a column detection model to identify table structures, a header classification model to categorize columns, and a cell identifier to extract data. This segmentation allows each model to specialize in a specific task, improving overall decoding accuracy while managing complexity through modular architecture
Solution Approach 2:
The patent applies preliminary action by first detecting columns and classifying headers before extracting cell data. The system prepares the image by identifying the structural framework (columns and headers) before performing the actual data extraction, which improves accuracy by establishing context for subsequent recognition tasks
2Loss of information
If traditional invoice recognition systems are used, then the system is simple, but the system fails to extract tracking information such as item types, costs, and quantities
Solution Approach 1:
The patent implements multi-functionality through AI models that can handle multiple types of information extraction simultaneously. The header classification model identifies various column types (item description, quantity, price, total), and the system can extract diverse tracking information including item types, costs, quantities, and other invoice data within a single unified framework
Solution Approach 2:
The patent introduces an intermediary layer of AI-based column detection and header classification between the image input and data extraction output. This intermediary layer analyzes the invoice structure, identifies relevant information fields, and guides the extraction process, ensuring comprehensive capture of tracking information while organizing it systematically
3Reliability
If AI-based models are used to detect columns and classify headers in invoice images, then the accuracy and efficacy of decoding is significantly improved, but the computational complexity and processing time increase
Solution Approach 1:
The patent segments the processing workflow into distinct sequential stages: column detection, header classification, and cell data extraction. This segmentation allows the system to process only relevant portions of the image at each stage, reducing unnecessary computational overhead while maintaining high reliability through specialized models for each task
Solution Approach 2:
The patent performs preliminary column detection and header classification to establish the invoice structure before extracting cell data. By preparing the structural framework in advance, the system avoids redundant processing during data extraction, improving overall efficiency while ensuring reliable identification of information fields
Data Source
AI summary
Methods, apparatus, and articles manufacture to decode documents based on images using artificial intelligence are disclosed. An example apparatus includes a model executor to input an image into a first artificial intelligence (AI)-based model to generate detected columns of text in the image; and input the image into a second AI-based model to classify the detected columns into categories; a cell identifier to identify rows or cells in the detected columns; and a report generator to: link information corresponding to the rows or cells in the detected columns with corresponding categories; and generating a report based on the linked information.