Graph Network Character Recognition Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing end-to-end character detection and recognition technologies in natural scenes face challenges in accurately recognizing characters in images, leading to recognition errors and unrecognized situations.
Innovation Solution
A character recognition method that determines semantic and position information of individual characters, constructs graph networks using this information, and calculates features to improve recognition accuracy by combining visual and semantic data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional end-to-end character detection and recognition technology is used, then the system is simple and fast, but recognition accuracy is poor leading to errors and unrecognized situations
Solution Approach 1:
The patent segments the character recognition task into multiple independent modules: character detection module, graph construction module, and graph decoding module. Each module processes specific aspects (detection, relationship modeling, recognition) separately, allowing optimized processing for each sub-task while improving overall accuracy without requiring complete system redesign
Solution Approach 2:
The patent introduces graph networks as an intermediary component between character detection and recognition. The graph structure mediates by establishing relationships between detected characters, enabling the system to leverage contextual and positional information, thereby improving recognition accuracy through an additional processing layer
2Measurement precision
If graph networks are constructed to improve recognition accuracy, then recognition precision improves, but computational complexity increases
Solution Approach 1:
The patent performs preliminary character detection and filtering before constructing graph networks. By pre-identifying candidate characters and their positions, the system reduces the input size for graph processing, thereby lowering computational complexity while maintaining the accuracy benefits of graph-based relationship modeling
Solution Approach 2:
The graph network construction focuses computational resources on local relationships between detected characters rather than processing the entire image globally. By building graphs only around detected character regions and their local relationships, the system achieves high recognition accuracy with reduced overall computational burden
Data Source
AI summary
A character recognition method, a character recognition apparatus, an electronic device and a computer readable storage medium are disclosed. The character recognition method includes: determining semantic information and first position information of each individual character recognized from an image; constructing a graph network according to the semantic information and the first position information of each individual character; and determining a character recognition result of the image according to a feature of each individual character calculated by the graph network.