Metal Structure Phase Classification via Automated Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for classifying phases in metal structures, such as high-strength steel plates, face challenges due to varying photographing conditions like etching conditions and photographer variability, leading to inaccurate and time-consuming manual segmentation and limited applicability of existing image analysis technologies.
Innovation Solution
A method and device for determining optimal photographing conditions using feature value calculation and phase classification models to accurately classify metal structure phases, even under varying conditions, by assigning labels to pixels in images and using a learned model for classification, thereby determining the best conditions for subsequent photography.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual segmentation is used to classify phases in metal structure images, then phase classification can be performed, but it requires a huge amount of time and causes large errors depending on the worker
Solution Approach 1:
The patent replaces manual mechanical segmentation with an automated image processing system that uses algorithms to classify phases. The system automatically processes images, calculates feature values, and determines photographing conditions without human intervention, thereby eliminating time consumption and worker-dependent variability while maintaining classification accuracy.
Solution Approach 2:
The system performs self-service by automatically determining optimal photographing conditions and executing phase classification without requiring skilled operators. The automated algorithm evaluates images, adjusts parameters, and performs segmentation independently, freeing the system from dependency on human expertise and time investment.
2Productivity
If binarization of luminance values is used for phase classification, then processing speed increases, but accurate classification cannot be performed when luminance value difference is not clear
Solution Approach 1:
The patent changes the parameter used for classification from simple luminance values to multiple feature values that capture various characteristics of metal phases. By transforming the image into feature space with multiple dimensions (including luminance, texture, shape, and color features), the system achieves both fast processing and accurate classification even when luminance differences are subtle.
Solution Approach 2:
The system transitions from one-dimensional luminance classification to multi-dimensional feature space classification. By extracting multiple features (luminance, texture, shape, color) and combining them in a higher-dimensional feature space, the patent enables accurate phase differentiation that cannot be achieved with luminance alone, while maintaining processing efficiency through automated feature extraction and classification algorithms.
3Adaptability or versatility
If photographing conditions are not standardized, then flexibility in observation is maintained, but phase classification accuracy varies depending on etching conditions and photographer
Solution Approach 1:
The patent performs preliminary determination of optimal photographing conditions before actual phase classification. By pre-establishing the best imaging parameters based on sample analysis, the system ensures consistent and accurate phase classification while maintaining the flexibility to adapt to different samples. The predetermined conditions serve as a reference that guides subsequent observations.
Solution Approach 2:
The system incorporates feedback mechanisms where the results of phase classification are used to refine and verify the determined photographing conditions. By continuously evaluating classification accuracy and adjusting parameters accordingly, the system maintains both consistency in results and flexibility in adapting to different metal structures and observation requirements.
Data Source
AI summary
A photographing condition determining method includes: photographing a part of a metal structure of a metal material subjected to predetermined sample preparation under a predetermined photographing condition; assigning, to pixels corresponding to one or a plurality of predetermined phases of the metal structure, labels of respective phases for a photographed image; calculating one or more feature values for a pixel to which a label of one of the assigned phases; classifying the phases of the metal structure of the image by inputting a calculated feature value to a model, which has been learned in advance using feature values assigned with labels of respective phases as input and labels of the respective phases as output, and acquiring a label of a phase of a pixel corresponding to the input feature value; and determining a photographing condition when other parts of the metal structure are photographed based on a classification result.