Terahertz Imaging Algorithm Based On Compressed Sensing

Terahertz wave has physical properties such as safety,spectral resolution,penetration and high resolution.It can be used as a signal source for object imaging,providing a new imaging method for many areas such as non-destructive testing,biomedical imaging,security and military.Since most terahertz imaging systems use raster scanning,the imaging time is long and the amount of data collected is large,which is seriously restricted in the practical application.In order to improve the imaging speed,hardwares can be improved but it is more difficult and spends more equipment costs.From the software approach,a better imaging algorithm can be used,which is more cost-effective and less difficult compared to hardware improvements.In this thesis,the reflective THz imaging experiment is performed and THz images of a small PCB are obtained.The reconstruction algorithms of time-domain THz spectrum and images based on compressed sensing are proposed.The validity of algorithms is verified by Matlab.The main contents of this thesis are as follows:(1)The orthogonal matching pursuit(OMP)algorithm for reconstructing 1-D time-domain THz spectroscopy is proposed.The sparse representation of signals,general mathematical model of compressed sensing,design method of measurement matrix and reconstruction algorithm are studied.The basic principle and implementation process of OMP algorithm are described.The ideal 1-D signal and actual pulse THz time-domain spectrum are reconstructed.The results show that almost complete reconstruction is achieved by 20 % of the measurement data.In order to determine the range of measurements quickly,a certain calculation condition is proposed for selecting the minimum number of measurements.(2)The reflective THz imaging experiment is performed.The modules and functions of the THz imaging system are introduced,and the working principle,experimental preparation and imaging process of the experimental system are expounded.A small PCB is used as the sample,and a local PCB circular THz image with a resolution of 0.2 mm and diameter of 3 mm is obtained.According to the intensity of the signal and its corresponding colors,the surface structure and material properties of the corresponding PCB detection area are identified,which verifies the application potential of THz imaging in PCB surface detection.(3)The OMP algorithm based on discrete wavelet transform for reconstructing 2-D THz images is proposed.Symlet wavelet is selected as the sparse transformation base,the implementation flow of WOMP algorithm is described.The 256×256 pixel Lena grayscale image and PCB terahertz grayscale image are reconstructed.The results show that WOMP algorithm can achieve good reconstruction with a sampling rate of 0.6,which can save about 40 % of the measurement data.In theory,the imaging speed can be increased by 40 %.The median reconstruction and color enhancement are carried out to improve the smoothness and discernibility of reconstructed images.