Simulation Algorithm For Hyperspectral Image Of Space-based Typical Complex Background

Hyperspectral simulation images have important application value in the development of spacebased target detection system.Due to the high precision requirements of space-based target detection system,researchers need many hyperspectral images to verify the design index of system,lots of real hyperspectral images are hard to acquire for the huge cost,hyperspectral simulation images provide many experimental images for researchers in the design of hyperspectral image detector.In addition,space-based hyperspectral imaging detectors cannot image space objects outside the designed band.Hyperspectral simulation images can provide researchers with images of other bands of interest outside the designed band.In this paper,the key issue of hyperspectral image simulation algorithm for space-based typical complex background is solved: the types and distribution of ground objects.An endmember extraction algorithm based on spatial-spectral joint constraint and a hyperspectral image unmixing algorithm based on spatial constraint are proposed.According to the obvious spatial continuity in hyperspectral image,Endmember extraction algorithm based on spatial-spectral joint constraint added regularization with spectral compactibility and spatial smoothness of endmembers into nonnegative matrix factorization to construct the objective function.The optimal solution is solved by means of multiplicative update rule in nonnegative matrix factorization.The unmixing algorithm based on spatial constraint put spatial information,which represent local similarity and local difference between adjacent pixels in hyperspectral images,into sparse unmixing model in the form of weighted manifold regularization,so that the output abundance matrix of hyperspectral images is consistent with the spatial structure information of hyperspectral images,and the accuracy of understanding the mixing is improved.This paper proposes and implements the simulation algorithm for hyperspectral images of spacebased typical complex backgrounds.The reflectance image of ground objects is calculated based on the input multispectral/hyperspectral images;unmixing spectral library with lower dimension is constructed through endmember extraction of the reflectance image and match in original spectral library,then distribution of ground object of the reflectance image is calculated by unmixing spectral library.Finally,hyperspectral simulation images of interested bands are generated through atmospheric model and detector model.The hyperspectral image simulation algorithm is verified by using four typical complex background hyperspectral images under APEX and Landsat8 detectors.Experimental results show that the simulation images are similar to the original images.