Research On Space Target Typical Structures Recognition Of ISAR Image Based On Model

With the continuous development of radar technology,the Inverse Synthetic Aperture Radar(ISAR)imaging resolution has been greatly improved.For the general space targets with large size and composing of typical structures,the details of their srtuctures can be shown in the high-resolution ISAR images.Distinguishing the typical structures of the space targets with these details can effectively assist the monitoring and recognition of space targets.In order to realize the recognition of the typical structures,this paper develops the research on space targets typical structures recognition of ISAR image based on model by reference to the recognition thought of radar images based on model.During the research,some problems to be solved mainly include: the segmentation of typical structures,the pose estimation of typical structures,the features extraction of typical structures and reasonable matching method.This article aims at the problems above and conducts the following research content.Analyze the ISAR image features of space targets,and exploit the structures distribution characteristics to propose a segmentation method for the space targets typical structures based on CLEAN.Firstly,we segment the main body by an adaptive region growing method,and then utilize an iterative correlation clustering method to finish the subcomponents segmentation.Finally,the actual and simulated ISAR images verify the validity of the segmentation method.Establish the ISAR geometric projection model of the space target typical structure and propose a two-dimensional(2-D)pose estimation method based on linear features differential projection.This method mainly uses the symmetry of the typical structures and extract the axis feature from the ISAR image;Then build the 2-D differential projection of the axis between the actual ISAR image and simulated image generated with the estimated pose,which transforms the absolute errors to relative errors about the detection of axis projection and decreases their effects on the pose estimation effectively.Some simulation experiments are made to verify the valid of this method.Simplify the attributed scattering centers model and keep the main scattering features,which are estimated by the image domain decoupling method.Based on these features,we use the matching method between the point features and target region to distinguish the types of typical structures.If the matching result can't realize the recognition,we also bring in the similarity measurement of frequency dependent factor and get the final recognition results.The simulation results show that this recognition method has better angular stability and noise immunity.