Automatic Graph Data Extraction Using Deep Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for extracting data from bar charts and line charts are inefficient and imprecise, relying on semi-automatic techniques that require manual labeling and estimation, which are time-consuming and prone to errors.
Innovation Solution
A deep learning-based method for automatic data extraction from graphs, involving text box localization, element identification, and positional correlation verification, to accurately extract data points and values from bar charts, line charts, and mixed graphs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual labeling and estimation methods are used to extract data from graphs, then the extraction process can be performed with simple tools, but the efficiency and precision of data acquisition deteriorate
Solution Approach 1:
The patent replaces manual mechanical operations (labeling, estimating, measuring) with an automated computer vision system that uses image processing algorithms to detect graph elements, identify data points, and extract values automatically, thereby eliminating the time-consuming manual process while maintaining or improving precision
Solution Approach 2:
The system enables self-service data extraction by allowing the graph image itself to provide all necessary information through automated analysis, where the computer vision system independently identifies axes, labels, data points, and extracts values without requiring manual intervention or additional input from users
2Productivity
If automated image processing methods are used to locate and identify graph elements, then data extraction efficiency improves, but the complexity of the processing system increases
Solution Approach 1:
The patent segments the graph image processing into distinct functional modules: graph element detection (axes, labels, ticks), data point identification, value extraction, and result compilation. This modular segmentation allows each component to be optimized independently while working together to achieve high overall efficiency
Solution Approach 2:
The computer vision system is designed with universal capabilities to handle multiple types of graphs (bar charts, line charts, scatter plots) and various graph element configurations through a single integrated framework, reducing the need for multiple specialized tools and simplifying the overall system architecture
3Measurement precision
If deep learning methods are applied for text box localization and element identification, then the accuracy of data extraction improves, but the computational resources and processing time required increase
Solution Approach 1:
The system performs preliminary processing steps such as image preprocessing, edge detection, and feature extraction before applying deep learning models. This preliminary action reduces the complexity of the main recognition task, allowing deep learning to focus on higher-level pattern recognition and improving overall efficiency while maintaining accuracy
Solution Approach 2:
The patent implements a continuous processing pipeline where detected elements from one stage are immediately used in the next stage without interruption. The system maintains continuous optimization by using feedback from detection results to adjust processing parameters in real-time, ensuring efficient resource utilization while sustaining high accuracy throughout the extraction process
Data Source
AI summary
A method for automatic extraction of data from a graph, including text area locating and text box classification; locating of coordinate axes, and locating of the positions of hatch marks on the coordinate axes; legend locating and information extraction; extracting corresponding bar or polyline connected components according to legend color, and filtering and classification; determining key points on the X-axis and locating a corresponding X-axis label for each key point; locating key points of the bars and polyline according to the X-axis key points, determining labeled numerical text boxes that correspond to the key points, and identifying the numerical text; calculating a corresponding value for each pixel, and estimating corresponding values of the key points of the bars or polyline; determining a final result according to a difference between the estimated values and the recognized labeled values.