Optical Character Recognition Rectification via Iterative Movement Profiles

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Hand-held scanners using line cameras face distortions in image data due to uncontrolled manual movement relative to the scanning frequency, leading to high data processing efforts for rectification and lack of optimization in movement profiles for character recognition.

Innovation Solution

A method that rectifies image data using predefined and iteratively optimized movement profiles, determining a global confidence value to maximize character recognition accuracy, reducing data processing effort and improving recognition reliability.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If image data is rectified using conventional methods (mechanical/optical detection or vector-based pattern recognition), then distortion is corrected, but data processing effort increases significantly

Engineering Contradiction:
Improveimage rectification accuracyVSAvoiddata processing effort
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent changes the approach from complex image analysis to simple parameter-based rectification. By detecting movement parameters (position, speed, acceleration) of the scanning device and applying corresponding transformation parameters to the image data, the system achieves accurate rectification with minimal processing effort. This transforms the problem from pattern recognition to parameter transformation.

Inventive Principle:
Principle #35Parameter changes

2Ease of operation

If manual uncontrolled movement is used for scanning, then ease of operation is improved, but image distortion occurs due to asynchrony with scanning frequency

Engineering Contradiction:
Improvemanual scanning convenienceVSAvoidimage data accuracy
Core Design Contradiction:
Ease of operationVSManufacturing precision

Solution Approach 1:

The system continuously detects the actual movement of the scanning device during manual operation and uses this feedback to dynamically adjust the image rectification parameters. By measuring position, speed, and acceleration in real-time and applying corresponding corrections, the system maintains high accuracy despite uncontrolled manual movement.

Inventive Principle:
Principle #23Feedback

3Measurement precision

If a priori information about character height and aspect ratio is used for rectification, then character recognition is improved, but the method fails with distorted or unfamiliar fonts

Engineering Contradiction:
Improvecharacter recognition accuracyVSAvoidfont compatibility
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent extracts the movement characteristics from the scanning process itself rather than relying on a priori information about character properties. By detecting the actual movement parameters during scanning and using these to rectify the image, the system adapts to any document type, font, or layout without requiring pre-defined character specifications.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP2338130B1Method for automatic optical character recognition, computer program product, data processing system, and scanner
Publication Date: 2014.04.02 BUNDESDRUCKEREI GMBH
  • EP2338130B1 patent drawingFigure 1
  • EP2338130B1 patent drawingFigure 2
  • EP2338130B1 patent drawingFigure 3

AI summary

The invention relates to a method for automatically recognizing optical characters in a document (122), comprising the following steps: at least one section (128) of the document containing characters is scanned by means of a linescan camera (102), an uncontrolled relative movement of the document and the linescan camera being performed such that image data (110) of at least the section is recognized; the image data is rectified with the help of a predefined movement profile (106) of a relative movement of the document and the linescan camera; characters in the rectified image data (112) are identified, a global reliability value for the correctness of the identified characters being determined; the movement profile is iteratively modified, the image data being rectified by means of the modified movement profile, and the characters being identified during each iteration. The iterative modification is made following an optimization process in order to maximize the global reliability value.