SEM Fatigue Striation Analysis with Radon and Spectral Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computer vision techniques struggle to robustly measure fatigue striation properties in scanning electron microscope images due to large appearance variations and insufficient training data, making it challenging to automate fatigue damage assessment in parts like aircraft components.
Innovation Solution
A computer-implemented method using image processing techniques such as Gaussian filtering, histogram equalization, and Radon transform, combined with spectral analysis, to enhance and analyze fatigue striations in SEM images, determining striation properties like angle and density, and assess fatigue damage by thresholding peak values in power spectra.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If traditional computer vision techniques are used to measure fatigue striation properties, then the measurement process can be automated, but the measurement robustness deteriorates due to large appearance variations and insufficient training data
Solution Approach 1:
The patent replaces traditional machine learning-based computer vision systems with a physics-based signal processing pipeline. Instead of using trained neural networks that require large datasets, the invention uses deterministic image processing operations (autocorrelation, Radon transform, spectral analysis) that are inherently more robust to appearance variations and do not require training data, thereby maintaining automation while improving measurement reliability
Solution Approach 2:
The patent transforms the image analysis problem from direct feature extraction to spectral domain analysis. By applying autocorrelation to enhance periodicity, then Radon transform to convert spatial information to frequency-space, and finally spectral analysis to identify dominant frequencies, the system extracts striation properties (density, orientation) as spectral parameters that are invariant to appearance variations, resolving the contradiction between automation and robustness
2Measurement precision
If manual measurement methods are used for fatigue striation properties, then measurement accuracy can be maintained, but productivity deteriorates due to time-consuming manual analysis
Solution Approach 1:
The patent replaces manual visual inspection and measurement with an automated signal processing system that applies mathematical transforms (autocorrelation, Radon, Fourier) to extract striation properties. This substitution maintains measurement precision by using objective mathematical criteria to identify striation characteristics while dramatically increasing productivity through automation, eliminating the time-consuming nature of manual analysis
Solution Approach 2:
The patent enables the measurement system to automatically determine striation properties without human intervention. The processing pipeline self-adjusts by applying sequential operations (autocorrelation to enhance patterns, Radon transform to analyze orientations, spectral analysis to extract frequencies) and automatically identifies striation density and orientation from the power spectrum, achieving both high precision and high productivity
3Measurement precision
If complex image processing operations are applied to enhance striation patterns, then measurement precision improves, but device complexity increases
Solution Approach 1:
The patent segments the image processing task into distinct sequential stages: autocorrelation computation to enhance periodic structures, Radon transform to convert spatial orientations to frequency representations, and spectral analysis to identify dominant frequencies. Each stage performs a specific function that builds upon the previous stage, improving measurement precision through systematic pattern enhancement while keeping individual processing modules relatively simple and well-defined
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In order to improve the measurement or assessment of fatigue damage to parts a computer-implemented method for measuring striation properties of fatigue striations (26) on a sample surface of a part (12) is proposed. A sample surface (20) is imaged using a scanning electron microscope so as to obtain a sample image (24) potentially containing the fatigue striations (26). A sample image patch (28) potentially containing fatigue striations (26) is selected from the sample image (24) for further processing. After normalizing the sample image patch (28) and enhancing line-like regular structures contained in the sample image patch (28) the resulting normalized image patch (34) is autocorrelated, Radon transformed and spectrally analyzed. The resulting power spectrum (42) of the transformed image patch (38) contains information about the striation properties of the fatigue striations (26) contained in the sample image (24), if any fatigue striations (26) are present. Furthermore, a system for performing the method is proposed.