Research Of 3D Temporal Speckle Pattern Interferometry

Because of its non-contact,high precision and high sensitivity advantages,electronic speckle pattern interferometry is widely used in industrial engineering,quality detection,aerospace,railway traffic,biopathology etc.To achieve 3D measurement and obtain displacements of objects in real time,this paper designs and builds a set of 3D speckle pattern interferometry measurement system.What's more,one denoising algorithm based on CNN net is also processed.Based on the principle of temporal speckle interferometry,this system can make dynamic measurement.With the CNN net,denoised images can be reconstructed conveniently.This system is different from the existed scheme and is the basis of instrumentation:The three-wavelength fiber-coupling laser is used to couple the three wavelengths of red,green and blue to produce the speckle patterns which finally are detected by CCD.The out-of-plane and on-plane measurements are optimized with the cage structure.To improve the result,speckle fringe needs to be denoised.Convolution neural network,as a typical training net,is widely used in image processing,super-resolution reconstruction,etc.This technology effectively meets the needs of improving interferometry patterns after the measurement was done.In this paper,the training data sets are simulated by MATLAB and the CNN net is programmed by PYTHON.At last,the algorithm based on the wavelet transform is programmed to retrieve the phases which are caused by the displacements of the object measured.This system is highly flexible,compact,robust and noise free.The main works of the paper:1.Research on 3D temporal speckle interferometry technology.Combining optical principles with mechanical assembly,design a robust and complete device.2.A setup of 3D temporal interferometry measurement that is robust and compact is built to retrieve the object's displacements,and the relative errors are 2.3%,0.5%,and 1.1%.3.Training data sets simulated for the CNN network realize the noise removal,this net successfully turns the noisy fringe in the experiment into clean images.