Character Recognition Device Dynamic Width Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing character recognition devices struggle to accurately estimate standard character width in character strings, particularly when dealing with contacting characters and aspect ratio deformed characters, as they often rely on fixed or predetermined character widths, which fail to account for variations in character size and shape.
Innovation Solution
A device and method that acquires character string images, measures character widths, provisionally recognizes characters through a matching process, and determines a standard character width by averaging measured widths, excluding widths of combination characters and kanji radicals to prevent misidentification, and adjusts based on perpendicular width if necessary, allowing for accurate recognition of full-width and half-width characters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a fixed character width is assumed for character cutting, then the device complexity is reduced, but the recognition accuracy deteriorates when full-width and half-width characters are mixed
Solution Approach 1:
The patent applies dynamics by transitioning from fixed predetermined character widths to dynamic character width estimation. The system estimates character width based on the actual imaging results and character string characteristics, allowing the character width parameter to adapt dynamically to different scanning scenarios and character types (full-width, half-width, mixed), thereby resolving the contradiction between device simplicity and recognition accuracy.
Solution Approach 2:
The patent changes the character width parameter from a static predetermined value to a dynamically estimated value based on imaging results. By calculating character width from actual image data and adjusting it according to character type detection, the system achieves accurate character width measurement across different scanning conditions without significantly increasing device complexity.
2Adaptability or versatility
If predetermined character widths are prepared for full-width and half-width characters, then the adaptability to mixed character strings is improved, but the device cannot respond to aspect ratio deformed characters
Solution Approach 1:
The patent applies preliminary action by performing character width estimation and character type classification before actual character recognition. The system first estimates the character width from the character string image, classifies the character type (full-width, half-width, or deformed), and then adjusts the recognition process accordingly. This preliminary characterization enables the system to handle various character types including aspect ratio deformed characters with high reliability.
Solution Approach 2:
The patent achieves universality by creating a unified character width estimation mechanism that handles multiple character types (full-width, half-width, and aspect ratio deformed characters) through a single adaptive process. The system universally applies character width estimation based on imaging results and automatically adapts to different character types, eliminating the need for separate predetermined width settings for each character type.
3Device complexity
If the peak value of projection character width histogram is used as character width, then the measurement process is simplified, but contacting characters cause the peak value to represent double the actual character width
Solution Approach 1:
The patent applies feedback by using character width estimation results to guide subsequent character cutting and recognition processes. The estimated character width feeds back into the character segmentation stage, allowing the system to adjust cutting positions and strategies based on the estimated width. This feedback mechanism enables accurate character width measurement even for contacting characters by preventing double-width estimation errors.
Solution Approach 2:
The patent applies partial action by estimating character width from a subset of character width measurements rather than requiring complete histogram analysis. The system estimates character width based on selected reference characters or statistical measures from the character string, avoiding the complexity of full histogram processing while maintaining measurement accuracy for both single and contacting characters.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An object of the present invention is to provide a character recognition device, a character recognition program, and a character recognition method capable of correctly recognizing the character even in the character string including the contacting characters and the aspect ratio deformed characters. A character recognition device includes: a character string image acquiring unit for acquiring a character string image; a measurement character length acquiring section for acquiring a measurement character length of measuring a character length in a character string direction with respect to a plurality of characters from the character string image acquired by the character string image acquiring unit; a character provisional recognition section for provisionally recognizing the character within the acquired measurement character length through provisional matching process; a standard character length determining section for determining a standard character length based on the measurement character length of the provisionally recognized character; a character cutout section for cutting out all characters to be recognized from the character string image based on the determined standard character length; and a character recognition section for actually recognizing the cutout character through actual matching process.