Multi-Scale GAN for Single-Frame Fringe Phase Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current single-frame fringe pattern analysis methods, such as Fourier transform fringe image analysis, suffer from low accuracy and poor fidelity for complex surfaces, while multi-frame methods are inefficient and unsuitable for moving objects, necessitating a more effective single-frame method for high-precision phase information extraction.
Innovation Solution
A multi-scale generative adversarial network is employed, comprising a multi-scale image generator and discriminator, trained with a comprehensive loss function to output sine and cosine terms from a single fringe image, enabling efficient and accurate phase calculation using the arctangent function.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If Fourier transform fringe image analysis is used, then measurement efficiency is improved, but measurement precision deteriorates
Solution Approach 1:
The patent applies segmentation by dividing the frequency domain into multiple regions of interest (ROIs) and processing each region separately with different window functions. This allows the method to capture both global efficiency benefits and local precision requirements, resolving the contradiction between measurement efficiency and precision by treating different spatial frequencies differently.
Solution Approach 2:
The patent implements local quality by applying different window functions (e.g., Hamming, Hanning, Blackman) to different regions of the frequency spectrum based on their specific characteristics. High-frequency regions use windows optimized for noise suppression, while low-frequency regions use windows optimized for resolution, thereby achieving high precision without sacrificing overall measurement efficiency.
2Measurement precision
If windowed Fourier transform is used, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent manages complexity by systematically varying key parameters such as window function selection, ROI positioning, and filter coefficients based on the specific measurement requirements. This allows the system to achieve high precision through parameter optimization rather than through complex structural modifications, keeping the implementation relatively simple while maintaining high measurement accuracy.
3Measurement precision
If windowed Fourier transform is used, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent reduces calculation time through segmentation by dividing the frequency domain into multiple ROIs and processing each region independently with optimized parameters. This avoids the need to process the entire frequency spectrum with uniform high-precision settings, thereby achieving high overall precision with reduced total computation time compared to traditional windowed Fourier transform methods.
Data Source
AI summary
The invention discloses a single-frame fringe pattern analysis method based on multi-scale generative adversarial network. A multi-scale generative adversarial neural network model is constructed and a comprehensive loss function is applied. Next, training data are collected to train the multi-scale generative adversarial network. During the prediction, a fringe pattern is fed into the trained multi-scale network where the generator outputs the sine term, cosine term, and the modulation image of the input pattern. Finally, the arctangent function is applied to compute the phase. When the network is trained, the parameters of the network do not need to manually tune during the calculation. Since the input of the neural network is only a single fringe pattern, the invention provides an efficient and high-precision phase calculation method for moving objects.