Terahertz Time-domain Fast Imaging Technology Based On KPCA And SP

The terahertz wave is an electromagnetic wave with a wavelength between 3 mm and 30 ?m and a frequency between 0.1 THz and 10 THz(1 THz = 1012 Hz).Due to its special spectral position,terahertz radiation has both microwave characteristics and partial optical theory.Non-destructive testing and quality control can be realized by utilizing the properties of terahertz wave spectrum.THz pulse imaging has good spatial resolution and has many features that are different from traditional physical imaging fields such as microwave,nuclear magnetic and X-ray.For example,the wavelength is shorter than the radio frequency and the photon energy is very low,which can be widely used in the papermaking and polymer industries,the food industry,the pharmaceutical industry and the art protection.Terahertz pulse imaging is one of two ways of terahertz imaging(the other is terahertz continuous wave imaging).Compared with continuous wave imaging,pulse imaging can obtain time domain information and phase information of the target sample,and is high.The frequency band has a high spatial resolution.In contrast,continuous wave imaging,although imaging speed is fast,can not get the material composition of the sample,the resolution is not high.At present,time-domain pulse imaging systems still use point-by-point sampling imaging,and in recent years,with the development of compression sensing technology and edge detection technology.Under-sampling technology came into being,and the sampling method also changed.There are Z-type and S-type scanning points.This experiment uses S-scan and spectral dimensionality recognition algorithms.This topic is based on terahertz time-domain pulse imaging technology,focusing on rapid imaging in the microscopic field and image target recovery.At the same time,combined with spectral dimension reduction recognition,the traditional point-by-point imaging algorithm is improved.The main points of the paper are summarized as follows:1.Focus on the development and background of terahertz time-domain pulse imaging.The development status of each part of the terahertz time-domain spectral imagingsystem is introduced.Including optical theory,parametric sources,devices,image algorithms,etc.,as well as the latest application progress of pattern recognition algorithms in terahertz spectroscopy,and the latest advances in these fields to advance the development of time-domain pulse imaging,predicting the future development of this imaging model.Directions provide the basic theoretical basis and experimental conditions for this experiment.2.Detailed description of the system components of the experiment,including laser parameter performance,spectrometer optical path and working principle,and the functions of each part of the computer remote control terminal.The imaging experiment process of this chapter is expounded,including the introduction of Ginkgo biloba leaves and experimental objectives,experimental requirements and the conditions required for the corresponding comparative experiments.Details how the computer console handles imaging data.Provide basic data support for subsequent experiments3.The main is to do a comprehensive exposition of the basic terahertz spectral correlation algorithm,including support vector machines,clustering analysis and principal component analysis.The feasibility,advantages and disadvantages of the above three methods are discussed in combination with the specific problems in this paper.A spectral identification scheme based on principal component analysis was proposed for three experimental results.The effect of kernel function based principal component analysis(KPCA)in experiments is highlighted.The results of the sample spectrum after training were analyzed and decided to be used as part of the algorithm.4.The terahertz image problem is discussed from the theory of compressed sensing.It mainly involves the step frame of compressed sensing,and focuses on the image reconstruction technology.According to the equipment and samples and test requirements of this paper,three reconstruction algorithms are also selected:orthogonal matching tracking,subspace tracking,SL0.Through the comparison of the three experiments,it is concluded that the subspace tracking algorithm is very suitable as a carrier for the innovative application of this paper.We then combine the improved subspace tracking with the previous phase of the KPCA algorithm inhardware and software.The intelligent acquisition,real-time reconstruction,and spectral analysis are used to improve the under-sampling imaging based on the compressed sensing theory,and the traditional mechanical scanning and point-by-point imaging techniques are abandoned.By comparing the experimental results,it is found that the time required for the scheme and the quality of the generated image have achieved the expected results.5.Based on the results of the previous experiments,it will draw conclusions and predict the future development of terahertz.