| Three-dimensional reconstruction can reconstruct the three-dimensional structure information of the target,and has important theoretical research and engineering application value in unknown area exploration,topographic surveying and mapping,urban development planning,battlefield array detection,etc.Three-dimensional reconstruction based on optical images is easily affected by factors such as weather and distance;while three-dimensional reconstruction based on a single synthetic aperture radar(SAR)image often cannot obtain complete target information,resulting in poor reconstruction effects,and even failure to reconstruct a three-dimensional structure.Therefore,it is necessary to study the three-dimensional reconstruction method based on multi-view SAR images to obtain target information and complete the reconstruction alltime and all-weather.This thesis first studies the multi-view SAR imaging method,analyzes the SAR image scattering characteristics,studies the registration method suitable for SAR images,and then uses the multi-view SAR images as the basis to match the extracted feature points and estimate the geometric parameters of the target,And finally reverse the threedimensional structure of the target to generate a spatial point cloud.The main research contents are as follows:1.The spotlight mode synthetic aperture radar imaging method is studied.According to the spotlight mode SAR imaging geometric model,the echo generation formula of spotlight mode SAR is analyzed and deduced,the polar coordinate format algorithm is proposed to process the echo data,the specific method of subaperture division processing is analyzed,and the multi-view SAR images are obtained by spotlight mode.The images provide a data basis for subsequent registration and reconstruction methods.2.The multi-view SAR image registration method based on edge features is studied.Analyzing the scattering characteristics of SAR images,it is proposed to improve the phase consistency edge feature extraction method through threshold binarization,which is used to extract the prominent and stable edge features of SAR images.Then the FourierMellin algorithm uses the extracted edge features to calculate the transformation relationship between multi-view SAR images,and realizes the registration of multi-view SAR images.3.The method of suppressing interference and noise through window function filtering is studied.In the Fourier-Mellin algorithm process,the amplitude spectrum of the Fourier transform appears false peaks,which seriously affects the registration effect and makes the registration unable to be stable.The method of using window function filtering to suppress spectrum interference proposed in this paper can effectively suppress false peaks and achieve stable image registration.4.The method of feature extraction and matching of multi-view SAR images is studied.The original image sequence is smoothed by the two-dimensional Gaussian function,and the SIFT method is used to extract the features of images of different scales.Finally,the nearest neighbor algorithm is proposed to match the features to obtain the transformation relationship between the image sequences.5.The multi-view SAR 3D reconstruction method based on feature points is studied.The imaging geometric model is established,the transformation matrix between the imaging angle of view and the actual spatial position of the feature points is obtained according to the model,the spatial coordinates of the feature points are calculated,and the spatial point cloud is generated to realize accurate 3D reconstruction based on multiview SAR images.The method researched and proposed in this thesis has been verified by simulation and measured data.The results show that these methods can effectively realize the threedimensional reconstruction of multi-view SAR images... |
PDF Full Text Request