Text Recognition Rotating Picture Units for Inclined Text
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text recognition technologies struggle with accurately recognizing large-angle inclined text, as current methods fail to correctly rotate characters to a horizontal state, leading to reduced recognition success and accuracy.
Innovation Solution
A method and apparatus for recognizing text that involves acquiring a text image, determining text boxes, selecting a text box to be recognized, rotating the corresponding picture unit to a target posture, and performing text recognition to achieve accurate recognition of inclined text.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If text recognition is performed on large-angle inclined text without rotation, then the processing speed is maintained, but the recognition accuracy deteriorates
Solution Approach 1:
The patent applies preliminary action by rotating the picture unit to a target posture before performing text recognition. This preprocessing step ensures that inclined text is corrected to a horizontal state, allowing the recognition model to accurately process the text without being affected by rotation angles, thereby improving recognition accuracy while maintaining efficient processing
Solution Approach 2:
The patent segments the text recognition process into distinct steps: detecting text boxes, determining picture units, rotating picture units to target postures, and performing recognition on corrected units. This segmentation allows each step to be optimized independently, with the rotation step specifically addressing the inclination issue without affecting other processing stages
2Measurement precision
If the picture unit is rotated to a target posture, then the recognition accuracy of inclined text is improved, but the processing time increases
Solution Approach 1:
The patent changes the orientation parameter of the picture unit by rotating it to a target posture (typically horizontal alignment). This parameter transformation converts inclined text into a standard orientation that recognition models can process efficiently, improving accuracy without requiring complex angled recognition algorithms
Solution Approach 2:
By performing rotation as a preliminary action before recognition, the patent prepares the picture unit in advance in the optimal orientation. This allows the recognition model to operate on pre-corrected data, reducing the computational complexity during the actual recognition phase and minimizing overall processing time
Data Source
AI summary
Provided is a method for recognizing a text. The method includes: acquiring a text image; determining at least one text box in the text image, wherein each of the at least one text box corresponds to at least one word; determining, from the at least one text box, a text box to be recognized; determining a picture unit corresponding to the text box to be recognized in the text image; rotating the picture unit to a target posture; and determining a target recognition result by performing text recognition on the picture unit in the target posture.