Study On Registration Method Of Multi-view 3D ISAR Scatter Sets

The three-dimensional ISAR imaging algorithm can obtain the three-dimensional ISAR scatters set of the target.Compared with two-dimensional images,it can reflect the true size and shape of the target,thereby providing necessary information support for parameter extraction and target recognition.But limited by radar resolution,the single-view scatter set contains less features and information of the target,and it is difficult to effectively extract parameters of the target.The scatter sets of different perspectives are complementary,and more complete data can be obtained after registration,but there are also a large number of non-overlapping parts between the scatter sets of different perspectives,which makes the registration of scatter sets more difficult than the registration of 2D images..Because the 3D ISAR scatters set is sparse and there are few overlapped parts between scatter sets of different perspectives,the registration of scatter sets is difficult.To solve this problem,the thesis studied the registration method of multi-view 3D ISAR scatter sets.The main work can be summarized as follows:1.The three-dimensional ISAR imaging technology is studied in detail.The basic principles of 2D ISAR imaging and three-dimensional ISAR imaging algorithm based on the threedimensional swing platform are introduced.The three-axis rotation of the platform can provide three-dimensional resolution for the target.By combining the parameter estimation algorithm based on the fractional Fourier transform,the parameters of the echo signal can be accurately estimated,and the imaging result can be obtained.Three-dimensional ISAR imaging data of three models from different perspectives were obtained through simulations.2.Aiming at the problem that the feature points of the 3D ISAR scatter set are easy to gather and difficult to reflect the overall features,a multi-scale feature points extraction algorithm based on ranking curvedness of neighborhood points.After the algorithm performs quadric surface fitting at each point and extracts the curvedness,it is proposed to select points with curvedness that meet the ranking requirements in multiple neighborhood scales as feature points of the scatter set,so that the extracted feature points are more representative.Simulations verify that the feature points extracted by the algorithm are distributed more uniformly and can reflect the overall characteristics of the scatter set.3.Aiming at the problem that there are few overlapped parts between the scatter sets of multi-viewpoints,which are prone to get mismatching points,a matching points screening algorithm based on CRMS and DRMS is proposed.The algorithm first uses the distance root mean square error(DRMS)evaluation function to obtain a matching points which have "oneto-one" relationship,and then proposes to use the coordinate root mean squared error(CRMS)evaluation function to extract four pairs of matching points with the highest matching degree.The feasibility of the proposed algorithm is verified by simulations.4.Aiming at the problem that the refine registration algorithm is easy to trap in local optimum,an iterative closest points method based on adaptive threshold is proposed.The algorithm proposes to calculate the adaptive threshold based on the distance of the pair of nearest points between the scatter sets after each iteration,and extracts an effective corresponding point set based on the threshold.Then the corresponding point set is used for iterative registration and whether the iteration is converged.The multiple evaluation indexes verify that the proposed algorithm can obtain better registration results.On the basis of 3D ISAR imaging,aiming at the difficult registration of multi-view 3D ISAR scatter sets,this thesis proposes a registration method of 3D ISAR scatter sets based on feature points extraction,matching points screening and iterative transformation calculation based on adaptive threshold.Simulation results verify the feasibility of the proposed method.