Study On Spectral Imaging Technology Based On Compressive Sensing

The spectral imaging technology has been carried out for several decades,from the initial to the present has experienced several technological changes,which has become an essential part of remote sensing,aerospace exploration and other areas.The traditional imaging spectrometer's information processing flow is sampling,compress coding,transmission,decoding.With the rapid development of imaging spectroscopy and the increasing demand for remote sensing detection,the amount of data obtained by the actual system becomes more and more huge,which presents a huge challenge for data storage,computing,and transmission.On the other hand,as the information is redundant,it need to abandon the excess information,resulting in increase the data storage and calculate.A new type of spectral imaging system,which is based on compressive sensing,is a spectral imaging system which is sample and compress at the same time by using compressive sensing theory.It has been developed a lot in recent years.One important technique is based on coded aperture imaging system(CASSI),which uses the coding template or programmable spatial modulator replace the traditional imaging spectrometer's incident slit,coding the target before imaging,the resulting signal and the encoded form has a mathematical relationship,based on the relationship,a mathematical equation meet the compressive sensing was built,then using an appropriate reconstruction method to restore all the information.As the CASSI system has a very low sampling rate,for most of the signal,the signal recovery accuracy is not enough,so the sampling rate need to be improved.Increasing the amount of information by multiple exposures is an appropriate method.This paper is studying on the system.The main works and innovations of this paper are as follows:1.A multi-frame image based on dual-dispersion coded aperture imaging system is proposed.The single-frame computation imaging system lost a lot of information due to the extremely low sampling rate.Therefore,it is proposed to improve the sampling rate by using multi-frame images and finally improve the accuracy of data reconstruction.The single-dispersion coded aperture imaging system uses spectral dimension and spatial dimension's correlation,it has a simple structure,but the mining of data's characteristics is not sufficient,while the use of dual-dispersion structure can make full use of a higher correlation between the spectral correlations,which improve the recovery accuracy.2.Derive a more rigorous new mathematical model.In the original mathematical model of the coded aperture spectral imaging system,because the measurement matrix contains a large number of 0,and the RIP property is not fully satisfied while increasing the computational complexity.According to the coding aperture and the information corresponding relationship,a new mathematical model was established,which more stringent to meet the principle of compressive sensing principle,decomposing the whole system's large problems into a large number of sub-problems.When reconstruct,the pressure of calculation and the a huge amount of data is greatly reduced,and data processing can be more convenient3.The information reconstruction is carried out by using different sparse base and reconstruction algorithms,and the recovery results are evaluated by a variety of different data recovery evaluation methods.The four kinds of orthogonal bases and a redundant dictionary are used to sparse the signal,and the method of training the sparse dictionary by KSVD is introduced.The data reconstruction is carried out by using four different reconstruction algorithms.The conclusion of the feasibility of the system is evaluated by using a variety of spectral curve recovery evaluation criteria.