Table Recognition via Merging Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing table recognition technologies, especially those using deep learning schemes, face inefficiencies and high error rates when processing table pictures in formats like PDF scans and images, as they struggle to accurately detect and extract content from tables without clear borders.
Innovation Solution
A method and apparatus that detect tables by extracting merging features and direction features from candidate table recognition results, determining to-be-merged rows and their directions, and adjusting the recognition results to improve accuracy, utilizing modules for candidate result determination, to-be-merged row determination, and merging direction determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If deep learning schemes are used for table recognition, then automation is improved, but recognition accuracy deteriorates
Solution Approach 1:
The patent segments the table recognition process into multiple distinct modules: candidate table detection, merging feature extraction, to-be-merged row determination, merging direction determination, and result adjustment. Each module handles a specific aspect of the recognition task, allowing for more precise control and higher overall accuracy while maintaining full automation.
Solution Approach 2:
The patent performs preliminary actions by first detecting candidate tables and extracting merging features before final recognition. The system pre-processes the image by identifying potential table regions and their structural characteristics, then uses this preliminary information to guide the subsequent recognition process, improving both accuracy and efficiency.
2Measurement precision
If manual recognition is performed, then recognition accuracy is improved, but processing efficiency deteriorates
Solution Approach 1:
The patent implements self-service by creating an automated system that performs table recognition without human intervention. The system extracts merging features, determines to-be-merged rows and their directions, and adjusts recognition results automatically, eliminating the need for manual recognition while achieving high accuracy through its multi-module architecture.
Solution Approach 2:
The patent incorporates feedback mechanisms where the system continuously refines its recognition results. By adjusting candidate table recognition results based on extracted merging features and determined merging directions, the system learns from its own outputs and improves accuracy automatically, combining the precision of manual review with the speed of automation.
3Speed
If simple detection methods are used, then processing speed is improved, but recognition accuracy deteriorates
Solution Approach 1:
The patent performs preliminary detection of candidate tables and extraction of merging features before detailed recognition. This preliminary action allows the system to quickly identify potential table regions and their structural characteristics, then focus computational resources on refining the recognition of these pre-identified regions, achieving both speed and accuracy.
Solution Approach 2:
The patent segments the processing into fast preliminary detection of candidate tables followed by more detailed analysis of merging features and row structures. This segmentation allows the system to maintain high processing speed through efficient candidate identification while achieving high accuracy through subsequent detailed analysis of the segmented components.
Data Source
AI summary
Embodiments of the present disclosure relate to a method and apparatus for recognizing a table, a device, and a medium. An embodiment of the method can include: detecting a table on a target picture, to obtain a candidate table recognition result; extracting a merging feature of the candidate table recognition result, and determining a to-be-merged row in the candidate table recognition result based on the merging feature; extracting a direction feature of the to-be-merged row, and determining a merging direction of the to-be-merged row based on the direction feature; and adjusting the candidate table recognition result based on the to-be-merged row and the merging direction of the to-be-merged row, to obtain a target table recognition result.