Research On The Spectral Imaging Based On Compressive Sensing

Spectral imaging technology is the combination of photography and spectroscopy, which possesses the resolution capability of not only space but also spectrum. This technology can achieve qualitative analysis, quantitative analysis and localization analysis, so it has been widely used in environmental remote sensing, military object identification, industry monitoring, biomedical imaging and so on. A spectral imager produces a two-dimensional spatial array of vectors which represent the spectral information for the corresponding spatial locations. The addition of spectral information goes with huge amounts of data, which are quite difficult to be processed, stored and transmitted. So it is very necessary to look out for an effective method of compression and sampling.Compressive sensing is a fast-emerging approach recently to simultaneous sensing and compression that enables a more efficient way for information acquisition. In this thesis, compressive sensing is introduced to the field of spectral imaging in order to solve the problem of huge amounts of data. With a distinct sensing modality from traditional imaging, the imaging system has advantages such as compact structure, noise immunity, not requiring mechanical scanning and so on.In this thesis the research status of spectral imaging and fundamentals of compressive sensing are introduced firstly. A multispectral imaging system based on compressive sensing is designed in the experimental platform. A transmission liquid crystal light valve is employed to implement the linear projection, and the linear array detector are used to record the intensity of modulated spectrum, the spectral images of the object are acquired combined with the reconstruction theory of compressive sensing. Several factors which can impair the imaging quality are discussed. It is indicate that due to the single-point detection and reconstitution algorithm, there exist mutual effects among different areas. Then the imaging scheme is applied to the fluorescence microscopic imaging successfully, a data preprocessing named intensity normalization is adopted to eliminate the influence of fluorescence decay during imaging process. Based on the fact that spectral data is compressible, a new image reconstruction scheme including coding in the spectral dimension is proposed. Signal with intensity modulation is dispersed, a second spatial light modulator is used to realize the spectral encoding together with spatial encoding. Then the scheme is simulated, verifying that the new method works well and the single-point spectral imager scheme is possible.