| Terahertz wave(THz)refers to the electromagnetic wave with a frequency of 0.1-10THz(corresponding to a wavelength of 3mm-0.03mm),which is located between the microwave band and the infrared wave segment in the electromagnetic spectrum,and lies in a special position of transition from electronics to optics.THz wave imaging technology has been widely used in biomedicine,nondestructive testing and military security because of its unique strong penetration and low energy level characters.However,the THz imaging technology developed up to now still has the problems of slow imaging speed,incoherent detectors cannot collect phase information and low imaging resolution.These problems have become one of the bottleneck problems in the further development of THz wave imaging technology.Based on these problems,this paper relies on the National Key R&D Program,focus on the computational imaging technology,solve the problems and limitations in the continuous THz imaging technology,and develop efficient and high-quality computational imaging methods for THz.Some related problems in the continuous wave THz imaging are deeply studied and solved.Calculates recovery of compressed sensing imaging technology,phase imaging technology and image reconstruction based on complex neural network technology is applied to the continuous THz imaging,based on the unique characteristics of THz wave is put forward,to face the actual application of real time efficiency,at the same time to obtain the quantitative amplitude and phase information of the THz imaging algorithm and high resolution,The imaging mechanism and experimental results of these algorithms are analyzed and compared.The main research work and innovations are as follows:1.Aiming at the problem of slow imaging speed that caused by point-by-point scanning in traditional continuous THz wave imaging,a compressed sensing imaging method with optimized data sorting is proposed,and the key problems in the application of compressed sensing method in THz band are studied.With the demonstration of simulation and experiment results,the factors affecting the performance of THz compressed sensing imaging system are studied,and corresponding improvement methods are proposed.In the aspect of lens selection of optical path,the using of long focal length and large numerical aperture lens can improve the SNR of the imaging system.On the structure design and implementation of the measurement matrix,the design suggestion of matrix element size is discussed,and the feasibility of shifting metal mask matrix is verified by experiments.In terms of the selection of measurement numbers,the proposal of using compressed sensing imaging sampling ratio in THz imaging based on different application scenarios is suggested.A continuous THz wave single pixel imaging system is designed and built according to these criteria.Secondly,an optimization algorithm is proposed to reduce the time of image reconstruction.This method uses the statistical ranking of the measured values and data selection to reduce the reconstruction time.The feasibility of this method is verified by simulation and experiment.Experimental results show that the image reconstruction performance of SCS-4 algorithm with optimized data processing is better than that of the traditional sequential sampling compressed sensing algorithm.The proposed method can reconstruct better image quality when the sampling rate is 25%,the imaging time was reduced by three quarters compared to point-by-point scanning,the average MSE of images recovered by SCS-4 algorithm was reduced by 32% compared with CS algorithm.2.Aiming at the problem that THz incoherent detectors cannot obtain the phase information of THz wave signals,a continuous THz wave phase recovery imaging method based on unequal spacing measurement is proposed.In the conventional multiple-plane phase retrieval method,the convergence speed due to wave propagations and measures with equal interval distance is slow and leads to stagnation.To address this drawback,we propose a nonlinear unequal spaced measurement scheme in which the interval space between adjacent measurement planes is gradually increasing,it can significantly increase the diversity of the intensity with the less number of required images.The feasibility analysis of the performance improvement of the proposed method is given by system-level model simulation.The effectiveness of the proposed method is verified from the aspects of parameter selection,anti-noise performance,and redundancy of recording plane information.The speed of iterative operation can be increased to more than one time by using the proposed method.Finally,We have demonstrated the proposed UE-MPPR method with both numerical simulation and measured experiments.Compare to the conventional MPPR method the results show that the method in terms of convergence speed and noise suppression,was superior to the traditional spacing,such as plane phase recovery method,using the recovery phase images can be directly converted to actual three-dimensional topography information of the object,The calculated results are consistent with the height information of the actual object.The designed THz lens-free phase recovery imaging experimental device has a simple structure,no need of optical lens and reference light,combined with the phase recovery method of multiple recording planes with nonlinearity spacing,which provides a new technical method for achieving high-quality and fast phase imaging in the THz field.3.Aiming at the problem of lower imaging resolution due to the long THz wave length,a THz imaging resolution enhancement method based on complex neural network is proposed.Based on the physical characteristics of THz waves,the complex convolutional neural network with phase information involved in the operation is constructed in this method,both the forward transmission network and backpropagation algorithm of the convolutional neural network are supported by complex operation.On this basis,a neural network model based on channel sensing and sub-pixel convolution is constructed.The data set is built by using THz imaging system of point spread function(PSF)and THz image degradation method.Both simulation and experimental are carried out to verified the enhancement effectiveness of the image resolution,with the clear reconstruction image of the 0.8 mm stripe board,the resolution of the original system is improved by 64%.Finally,the advantages of the complex neural network superior to the original real neural network are evaluated quantitively,and the results demonstrated the proposed complex neural network method can effectively improve the resolution of THz imaging and the quality of image reconstruction. |
PDF Full Text Request