Study On The Inversion Analysis Of Scattering Spectrum Detection Based Partial Least Squares

Identification inversion of the target plays a very important role in national defense,military, scientific experiments and agricultural production, particularly in aerospace, materials, energy, metallurgy. The common methods applied in detected inversion always influenced by background noise, anti-identification materials, that lead us can not realized the good inversion results.The scattering spectrum like the fingerprint of the object, which not only including the material information, surface morphology and size information of test target, but also have the strong anti-jamming capability. The purpose of this thesis is to develop an inversion method, which applied the scattering spectroscopy of targets to do the identification inversion.In this thesis, the analytical method is the principle of reflectance function superposition, and through the indoor experiment of scattering spectroscopy detection of the target to proof the principle, which provides a theoretical basis to the application of target reflectivity function in the process of inversion. The thesis introduced a multivariate statistical data analysis method called partial least squares regression method, which was applied in the process of the inversion analysis of scattering spectroscopy. Then base on the different pretreatment methods and experimental results, summarized the best processing methods of inversion data. The accuracy of the final inversion results was up to 4.6%, and shows a good result of the proportion of each component in the target area. This study provide new ideas for the inversion research of long range goal, and also provide the theoretical guidance and experiments reference for further research of scattering spectra.