Hologram Reconstruction Based On Neural Network

With the advantages of non-contact,high-resolution and real-time,digital holography is widely used in the fields of shape deformation measurement,biomedical cell imaging analysis,particle field measurement and so on.Digital holography technology combines traditional optical holography technology with photoelectric detection technology,and records and reproduces the real three-dimensional image of an object by means of interference principle and diffraction principle.For digital holography,through theoretical analysis,numerical simulation and experimental means,the feasibility and performance of deep neural network for target amplitude and phase reconstruction and phase unwrapping are studied.Firstly,the theory of digital off-axis holographic imaging is introduced.For digital holographic imaging technology,this paper proposes a convolution neural network model.The simulation experiment of digital off-axis holography is established by MATLAB software.A large number of holographic images are randomly generated as training set to train the convolution neural network,and the mathematical model of holographic reconstruction based on deep learning is obtained.Numerical simulation results show that the convolution neural network can directly reconstruct the target image from a digital hologram with high efficiency,small amount of calculation,good real-time performance and high reconstruction accuracy.Secondly,the deep learning reconstruction method of digital holography is studied and verified by experiments.A digital off-axis experimental platform is set up.The hologram of the target is collected and the hologram is recorded by CCD.The network model is trained.Finally,for phase unwrapping,a phase unwrapping model based on convolution neural network is established.The phase of the network is wrapped by MATLAB,and the phase diagram is generated by wrapping the network model.The simulation results show that the convolution neural network can well realize the phase unwrapping operation.Figure 39;Table 8;Reference 60...