Character Recognition Apparatus for Multi-Line Unstructured Text
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing character recognition systems struggle to process character strings without a known format, particularly when they are printed over multiple lines, as seen in images of objects like labels on terminals and cables.
Innovation Solution
A character recognition apparatus and method that detects character regions in an input image, determines their direction, recognizes the characters or character strings, and connects character regions with the same direction that are close to each other, allowing for the recognition of character strings over multiple lines without a known format.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If character recognition is performed on character strings printed over multiple lines without a known format, then the recognition accuracy of multi-line character strings is improved, but the complexity of the recognition system increases due to the need to connect character regions and determine directions
Solution Approach 1:
The patent segments the character recognition process into distinct stages: detecting individual character regions, determining the direction of each region, recognizing characters within regions, and finally connecting regions that belong to the same character string. This segmentation allows the system to handle complex multi-line character strings by breaking them down into manageable components while maintaining overall recognition accuracy.
Solution Approach 2:
The patent introduces directional information as an additional dimension for character region analysis. By determining the direction of each character region and using this directional attribute to connect regions, the system can accurately reconstruct character strings that span multiple lines and orientations without significantly increasing overall system complexity.
2Reliability
If character regions are connected based on direction and distance criteria, then the ability to recognize continuous character strings across lines is improved, but the computational processing time increases
Solution Approach 1:
The patent performs preliminary detection of character regions and determination of their directions before the actual character recognition and connection processes. By pre-processing and organizing character region information including directional attributes, the system reduces the computational burden during the connection phase, thereby maintaining character string continuity while minimizing additional processing time.
3Adaptability or versatility
If the system processes character strings without referring to a format database, then the versatility of the recognition system is improved, but the accuracy of recognizing structured formats like mail addresses decreases
Solution Approach 1:
The patent creates a universal character recognition system that can handle both structured formats (like mail addresses) and unstructured formats (like labels on various objects) using the same core process. By relying on directional analysis and spatial connection of character regions rather than format-specific rules, the system achieves versatility across different application scenarios while maintaining sufficient accuracy for both structured and unstructured text.
Data Source
AI summary
The computing circuit detects a plurality of character regions in the input image, each of the plurality of character regions including a character or a character string made of a plurality of characters. The computing circuit determines a direction of the character or the character string in each of the character regions. The computing circuit recognizes the character or the character string in each of the character regions. The computing circuit generates a connected region by connecting at least two character regions including the character(s) or the character string(s) having a same direction, the at least two character regions being close to each other at a distance less than a threshold. The computing circuit connects the character(s) or the character string(s) included in the connected region to each other.