Adaptive Image Deformation for Automated Measurement Variations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional automated image measurement systems fail to handle variations in images due to changes in manufacturing processes and imaging conditions, leading to erroneous or incomplete measurements, which require manual intervention by engineers, resulting in errors and inefficiencies.
Innovation Solution
A method and system that use machine learning models trained with original and synthetic images to automatically determine image processing algorithms for measuring attributes of images, handling variations by generating synthetic images through targeted deformation based on manufacturing parameters, thereby reducing the need for manual measurements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If a setup is created based on a first image for automated measurement, then measurement automation is achieved, but the system fails to handle variations in remaining images due to manufacturing process changes and imaging conditions
Solution Approach 1:
The system performs preliminary actions by creating multiple synthetic images with anticipated variations before actual measurement occurs. These synthetic images are generated by deforming features of a reference image to simulate different manufacturing conditions, allowing the measurement system to be pre-trained and adapted to handle variations without requiring manual intervention when variations occur.
Solution Approach 2:
The system creates copies of the reference image through synthetic image generation, producing multiple variations that simulate different manufacturing conditions. These synthetic copies are used to train the measurement system, enabling it to recognize and measure attributes across diverse real-world variations without requiring physical samples of every possible variation.
2Measurement precision
If manual measurement is performed by engineers for images with variations, then measurement accuracy may be maintained, but engineer time increases and errors still occur
Solution Approach 1:
The system enables self-service by allowing the measurement system to automatically adapt and handle variations independently. Through synthetic image training, the automated system learns to recognize and measure varied structures without requiring engineer intervention, thereby maintaining measurement precision while eliminating the time cost of manual measurements.
3Adaptability or versatility
If synthetic images are generated through targeted deformation, then training data coverage is improved, but computational complexity increases
Solution Approach 1:
The system applies local quality by performing targeted deformation only on specific features that are expected to vary based on manufacturing parameters, rather than randomly deforming entire images. This focused approach generates sufficient training data coverage for critical variations while minimizing unnecessary computational complexity.
Data Source
AI summary
A method includes determining a plurality of features of a first original image of a first product that are expected to be different for one or more products to be produced via manufacturing parameters of a manufacturing process compared to the first product. The method further includes adjusting one or more of the plurality of features of the first original image to generate a first synthetic image. The method further includes providing a plurality of images including the first original image and the first synthetic image to train a machine learning model to generate a trained machine learning model configured to generate output associated with updating the manufacturing parameters of the manufacturing process.