Research On High-Resolution Image Reconstruction Technology Of Micro Satellite Constellation

With the high resolution remote sensing image is widely used in military and civilian areas, the requirement for image resolution is getting higher and higher. How to obtain high resolution remote sensing images with little cost has recently become a research focus in the field of space and aerial remote sensing. This paper presents a method of improving resolution of images obtained by the micro satellite using image registration and reconstruction technology. This paper includes the following contents:Firstly, the theoretical basis of imaging and the acquisition of the micro-satellite images are studied. Based on the analysis of the imaging technology, the feasibility and effectiveness of the method proposed in this paper are explained. According to the characteristics of a single satellite and the relation between the multi satellites, two methods of obtaining a sequence of images from micro satellite are presented.Secondly, several registration algorithms based on sub-pixel are compared. Four registration methods are adopted and compared through simulation experiment, which are based on the maximum mutual information, the SIFT, the SURF and the Keren. The simulation results proved that the registration error of the Keren is less than 0.1 pixels, which is most suitable for the processing of micro-satellite images.Then, two kinds of reconstruction methods are proposed. After the integerpixel translation, there are only sub-pixel displacements between those lowresolution images. To improve the spatial reso lution of remote sensing image, two kinds of reconstruction methods are proposed, which are sequential weighted method and iterative decomposition method.Finally, the qualities of images are evaluated and experiments are implemented. In order to simulate images obtained by micro satellite, the simulation experiments, laboratory experiments and field experiments are designed. Then, this paper uses Keren registration method and two kinds of reconstruction methods to process those images. At last, peak signal to noise ratio, information entropy and some other methods are used to evaluate the qualities of the images. The results proved that the space resolutions of the images are increased by nearly two times and this method can effectively improve the spatial resolution of images.