Unified OCR System for Imaged Media Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current OCR technologies face challenges in recognizing imaged information-bearing media, including the need for separate applications for different types, insufficient handling of rotated images, background text interference, and trapezoidal distortion leading to character recognition failures.
Innovation Solution
A method and apparatus that acquire and correct images of imaged information-bearing media using target detection and correction, text direction detection, and a CRNN text recognition network model to classify and archive text content, addressing issues of rotation, background interference, and distortion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If separate applications are used for different types of imaged information-bearing medium recognition, then recognition accuracy for specific types is improved, but device complexity and ease of operation deteriorate due to needing to switch among multiple applications
Solution Approach 1:
The patent combines multiple type-specific recognition functions into a single unified recognition system. The system automatically detects the type of imaged information-bearing medium and applies appropriate recognition algorithms, eliminating the need for separate applications while maintaining recognition accuracy across different types such as business cards, ID cards, and waybills.
Solution Approach 2:
The recognition system is designed with multi-functionality to handle various types of imaged information-bearing media through a single application. It includes type detection capabilities and adaptive recognition algorithms that can process different medium types universally, providing both versatility and ease of operation.
2Measurement precision
If automatic rotation correction is implemented, then recognition accuracy for rotated images is improved, but processing time increases
Solution Approach 1:
The system performs preliminary rotation detection and correction before the main recognition process. By detecting the rotation angle in advance and pre-correcting the image orientation, it ensures accurate recognition of rotated images while minimizing processing time during the actual recognition phase.
Solution Approach 2:
The patent replaces traditional mechanical or manual rotation correction methods with automated image processing algorithms. Using computer vision techniques to detect text orientation and automatically rotate images eliminates manual intervention and optimizes processing efficiency.
3Measurement precision
If background text filtering is implemented, then recognition accuracy is improved by avoiding false positives, but processing complexity increases
Solution Approach 1:
The system applies local quality analysis by focusing recognition on specific regions of interest within the image. It identifies and prioritizes text areas that match the expected characteristics of the target medium type, while filtering out background text through localized analysis rather than processing the entire image uniformly.
Solution Approach 2:
The patent extracts and isolates the relevant text content from the background by detecting text regions and separating them from non-relevant areas. This extraction process removes background interference while maintaining the integrity of the target text for accurate recognition.
4Measurement precision
If perspective transformation is applied to correct trapezoidal distortion, then recognition accuracy is improved, but computational requirements and processing time increase
Solution Approach 1:
The system dynamically adjusts the degree of perspective transformation based on the detected distortion level. By changing the transformation parameters adaptively rather than applying fixed maximum correction, it achieves sufficient distortion correction while reducing unnecessary computational overhead for images with minor distortion.
Data Source
AI summary
A method and apparatus for recognizing an imaged information-bearing medium, a computer-readable storage device and a computer device are provided. The method comprising: acquiring a first image of the imaged information-bearing medium; performing text recognition on the first image to acquire a text content of the imaged information-bearing medium; classifying the imaged information-bearing medium to acquire a type of the imaged information-bearing medium; and archiving the text content according to the type.