Image Region Correction for Training Data Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image inspection systems face challenges in accurately reflecting brightness and contrast variations within images, leading to low expression accuracy in secondary training images generated for machine learning.
Innovation Solution
An information processing system that divides captured images into regions, creates correction equations for each region based on feature amounts, and generates inference images using these correction equations, which are then used as training data for machine learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a classified image is learned as training data based on feature amounts extracted from the entire image, then machine learning can be performed, but the feature amounts do not accurately reflect brightness and contrast variations in different regions, resulting in low expression accuracy of training images
Solution Approach 1:
The captured image is divided into multiple divided regions, and feature amounts are extracted for each divided region separately. This segmentation allows the system to capture local brightness and contrast variations in each region, significantly improving the expression accuracy of training images compared to extracting feature amounts from the entire image as a whole.
Solution Approach 2:
Correction equations are created for each divided region based on its specific feature amounts, enabling localized brightness and contrast correction. This local quality approach ensures that each region's unique characteristics are preserved and accurately represented in the inference images, rather than applying a uniform correction to the entire image.
2Reliability
If the number of training data is increased by secondarily creating images with defects, then classification accuracy is enhanced, but the training images have low expression accuracy when brightness and contrast differ between regions
Solution Approach 1:
By dividing the image into multiple regions and extracting feature amounts for each region, the system accurately captures local characteristics. This segmentation ensures that when training images are secondarily created, they maintain high expression accuracy by preserving the unique brightness and contrast properties of each divided region.
Solution Approach 2:
The system creates correction equations that modify brightness and contrast parameters for each divided region based on its specific feature amounts. This parameter change approach allows the generation of multiple inference images with varied brightness and contrast levels, providing diverse and accurate training data that improves classification accuracy.
Data Source
AI summary
Provided is an information processing system, including: a division unit configured to divide a captured image; a correction equation creation unit configured to create a correction equation for each divided region of the divided captured image based on a feature amount expressing an image of the divided region; an inference image creation unit configured to create an inference image expressed by the feature amount for each divided region according to the correction equation for each divided region; and a learning unit configured to execute machine learning on a learning model in which the inference image is used as training data, and the captured image is used as input data.