Character Detection Apparatus Using Priority-Based Scheme Integration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing character detection methods fail to accurately detect character strings in images due to variations in lighting conditions and similarities between characters and backgrounds, and mechanical learning-based detectors struggle with unfamiliar logos or handwriting styles.
Innovation Solution
A character detection apparatus that includes a feature extractor, a priority determiner, and an integrator, which extracts feature values from images, determines the priority of different detection schemes based on accuracy, and integrates text line candidates from multiple schemes to select the most accurate detection, using a combination of connected component and mechanical learning-based approaches.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a connected component method is used to detect character strings, then simple implementation is achieved, but detection accuracy deteriorates when characters have similar colors to background or when lighting conditions vary
Solution Approach 1:
The patent segments the character detection task into multiple independent detection schemes (connected component method, mechanical learning method, and other alternative schemes). Each scheme operates independently to detect character strings, and their results are subsequently integrated through a voting mechanism. This segmentation allows the system to leverage the strengths of different methods while mitigating their individual weaknesses, particularly improving accuracy for challenging cases like similar colors or varying lighting conditions.
2Adaptability or versatility
If a mechanical learning-based detector is used, then adaptability to learned patterns is improved, but detection accuracy deteriorates for unfamiliar logos, regular handwriting, or ornamental writing
Solution Approach 1:
The patent merges multiple character detection schemes including the mechanical learning method with alternative detection methods such as the connected component method. By combining these different approaches and integrating their results through a voting mechanism, the system achieves better overall detection accuracy. The integration allows the system to leverage the pattern recognition strength of mechanical learning while compensating for its weakness in handling unfamiliar patterns through the complementary strengths of other detection schemes.
3Measurement precision
If multiple character detection schemes are integrated, then detection accuracy is improved, but device complexity increases
Solution Approach 1:
The patent implements a self-service mechanism where the system automatically determines the optimal detection scheme or integration strategy based on the characteristics of the input image. The voting mechanism and automatic integration process eliminate the need for manual configuration or complex user intervention, allowing the system to adaptively select and combine detection schemes. This self-service approach manages system complexity by automating the decision-making process rather than requiring manual control over multiple detection parameters.
Data Source
AI summary
According to one embodiment, a character detection apparatus includes a feature extractor, a determiner and an integrator. The feature extractor extracts a feature value of an image including character strings. The determiner determines each priority of a plurality of different character detection schemes in accordance with character detection accuracy with respect to an image region having a feature corresponding to the feature value. The integrator integrates text line candidates of the character detection schemes, and selects, as a text line, one of the text line candidates detected by the character detection scheme with the highest priority if a superimposition degree indicating a ratio of a superimposed region among the text line candidates is no less than a first threshold value.