The Study Of Accurate Registration And Related Technology For Spatial Objects

As the number of on-orbit spacecrafts grow, how to obtain their high-resolutionimages is becoming an important part of the space objects detection. Lucky imaging is apost-processing technology which can effectively reduce the effect of atmosphericturbulence on the image resolution. This technology greatly improves image qualitythrough special processes on the sequence of short exposure images of the objects,including image selection and registration and co-adding of the selected best images.When combined with adaptive optics, lucky imaging even gets images better than thattaken from the Hubble space telescope.This thesis is based on the processing procedure of lucky imaging. Firstly, theperformances of five commonly used evaluation functions applied to blurred and noisytest images are compared and analyzed through experiments, and the results show thatthe functions based on Sobel operator and Fourier frequency spectrum are more robustto noise and higher sensitive to blurring. Secondly, an image enhancement methodbased on steering kernel regression is improved through introducing Laplacian filterinto the steering kernel，which gives good results in denoising and enhanced display ofthe test images. Thirdly, an registration method based on phase correlation is improved,in which the rotation parameter of test images is obtained based on a two-step methodand the transition parameter is obtained based on matrix multiplication. Sub-pixelregistration accuracy is acheved with the improved method when images exist rigidtransformation, and the method is proved robust to noise and blurring as well. Finally,actual images are processed by use of lucky imaging technology. The results show thatthe lucky imaging technology can effectively improve the quality of images, and theprocessed images reveal more details and the noise is significantly reduced.