Table Image Recognition Using Set and Row-Column Determination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional table-image recognition techniques fail to correctly identify the structure of complicated tables with many elements other than character strings, such as images or figures, especially when elements are arranged across columns or rows, or when borders are unclear or light-colored, leading to incorrect allocation of elements to rows and columns.
Innovation Solution
A table-image recognition device and method that analyzes table images by extracting objects, determining pairs of objects as table components, and performing same-row and same-column determinations using machine learning models to accurately specify the structure of the table, even without clear borders, by treating combinations of elements as semantic elements and using character string content for determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional table-image recognition techniques are used, then simple tables with clear borders can be recognized, but complicated tables with many elements, unclear borders, or light-colored borders cannot be correctly recognized
Solution Approach 1:
The patent segments the table recognition task into multiple independent determination processes: set determination (identifying component groups), same-row determination (identifying row relationships), and same-column determination (identifying column relationships). Each process operates independently on specific features, allowing the system to handle complicated tables with various element types and border conditions without requiring a single complex recognition algorithm.
Solution Approach 2:
The patent transitions from traditional two-dimensional border-based recognition to a multi-dimensional approach by adding a third dimension of relationship analysis. Instead of only detecting borders and assigning elements to cells, the system determines relationships between elements across rows and columns, enabling accurate recognition of tables where elements span multiple rows or columns and where borders are unclear or light-colored.
2Ease of manufacture
If elements are detected separately and rectangular areas are specified, then element extraction is simple, but elements that should be treated as semantic units (such as image-character string combinations) are incorrectly allocated to different rows and columns
Solution Approach 1:
The patent performs preliminary set determination before row and column allocation. This preliminary action identifies which elements form semantic units (sets) that should be treated together, such as image-character string combinations. By establishing these set relationships in advance, the system ensures that elements belonging to the same semantic unit are correctly allocated to the same row and column, preventing the errors that would occur if elements were processed independently.
3Difficulty of detecting and measuring
If border detection is used to define rows and columns, then clear borders can be easily detected, but light-colored or unclear borders cannot be correctly processed
Solution Approach 1:
The patent introduces relationship determination as an intermediary process between element detection and row-column allocation. Instead of directly using border detection to assign elements to cells, the system first determines set relationships, same-row relationships, and same-column relationships among elements. This intermediary relationship analysis enables the system to infer row and column assignments even when borders are light-colored or unclear, by using the semantic relationships between elements as mediators to establish the table structure.
Data Source
AI summary
A table-image recognition device includes: an object extracting unit that extracts a plurality of objects included in a table; a set determination unit that determines whether or not every pair consisting of two objects selected from the plurality of objects is a set constituting a component specified by a column and a row of the table; a same-row determination unit that determines whether or not the objects of each pair share a same row; a same-column determination unit that determines whether or not the two objects of each pair share a same column; and a structure determining unit that determines a structure of the table by specifying the row and column to which each object belongs on the basis of the determination result.