Image Filter Generation for OCR Misrecognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Optical character recognition (OCR) systems trained through machine learning struggle with environmental fluctuations such as light changes and workpiece misalignment, leading to character misrecognition.
Innovation Solution
An image filter generation system that learns relationships between article image data and image filters using machine learning to infer optimal image filters for OCR, reducing misrecognition by generating trained models based on OCR score data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If predetermined high-reliability image filters are prepared alone and learning combinations are performed, then the device can infer optimal filter combinations for image processing, but the device cannot respond to environmental fluctuations such as light changes and workpiece misalignment, leading to character misrecognition
Solution Approach 1:
The system dynamically generates image filters based on the specific characteristics of each input image rather than using predetermined fixed filters. The filter generation unit creates customized filters by analyzing image properties such as noise levels, contrast, and character patterns, allowing the system to adapt to varying environmental conditions including light changes and workpiece misalignment.
Solution Approach 2:
The system changes the parameters of image filters based on the specific characteristics of the input image. The filter generation unit adjusts filter parameters such as kernel size, sigma values, and threshold levels according to the analyzed image properties, enabling the system to respond effectively to different environmental conditions and workpiece variations.
2Measurement precision
If machine learning is used to train image processing devices, then the devices can recognize characters in controlled environments, but they fail to maintain accuracy when environmental conditions change during actual operation
Solution Approach 1:
The system performs preliminary analysis of the input image characteristics before generating filters. The filter generation unit examines noise patterns, contrast levels, and other image properties in advance, then generates appropriate filters tailored to these specific characteristics, enabling accurate character recognition despite environmental variations.
Solution Approach 2:
The system uses OCR score data as feedback to evaluate the effectiveness of generated filters. The filter generation unit creates filters and evaluates their performance through OCR scoring, then refines filter parameters based on this feedback to optimize character recognition accuracy for the specific environmental conditions present.
Data Source
AI summary
A training data acquirer acquires training data including article image data, image-filter-related data indicating a combination of a plurality of image filters used for image processing of the article image data and a value of a parameter for each of the plurality of image filters, and optical character recognition (OCR) score data indicating a score of character recognition output through OCR when image processing is performed on the article image data using the image filters based on the image-filter-related data. A trained model generator generates a trained model indicating a relationship between the article image data, the image-filter-related data, and the OCR score data through machine learning using the training data.