Convolutional neural network for estimating physical parameters from Newton's rings

Towards a more practical analysis of Newton's rings using deep learning

As a typical form of optical fringes with a quadratic phase, Newton's ring patterns play an important role in spherical measurements and optical interferometry. A variety of methods have been used to analyze Newton's ring patterns. However, it is still rather challenging to fulfill the analysis. We present a deep-learning-based method to estimate the parameters of Newton's ring patterns and fulfill the analysis accordingly. The experimental results indicate the excellent accuracy, noise robustness, and demodulation efficiency of our method. It provides another applicable approach to analyzing Newton's ring patterns and brings insights into fringe analysis and interferometry-based measurements.

Concise fractional Fourier transform based on a non-uniform order searching method for estimating physical parameters from Newton's rings

Newton's ring pattern is very common in interferometry. By analyzing it, the physical parameters can be estimated, such as the curvature radius and the rings' center. However, parameter estimation from large images using fractional Fourier transform (FRFT) consumes considerable time. A concise FRFT based on a non-uniform order searching method is proposed to reduce the computational time without decreasing the accuracy. Experimental results show that the time of the proposed method is about 2.7 s, which is about 1/600 of that of the traditional FRFT-based method and 1/5 of that of the Fast FRFT-based method when processing 960×960 pixel images.