Опубликована: Дек. 30, 2023
Wavelength tuning interferometry is widely used in optical metrology order to obtain the phase information of sample. The obtained wrapped usually unwraps range [-π, π]. Therefore, unwrapping operation required right phase. But some factors, such as shift miscalibration, coupling error, and noise, always lower precision conventional shifting algorithm. To address kind problems, we proposed a deep learning method using convolutional neural network perform process by turning task into multiclass classification work 2N - 1 for generating training dataset. results indicated that not only can compensate miscalibration but also has strong robust denoise ability, which means outperformed other measurement algorithms.
Язык: Английский