TomoSAR Height Reconstruction For Buildings With Limited Observation

Height reconstruction of three-dimensional(3-D)building models in urban areas is a hot topic in remote sensing,photogrammetry,and computer vision.Synthetic Aperture Radar Tomography(Tomo SAR)is a remote sensing technique that extends the conventional two-dimensional(2-D)SAR imaging principle to three-dimensional(3-D)imaging.SAR has the ability to work all day due to the active emission of signals.Moreover,SAR is almost independent of weather conditions because of the use of microwaves in radar signals,which is a major advantage when compared to sensors in the visible or infrared spectrum.And Tomo SAR techniques introduce the idea of synthetic aperture to the elevation,which make it possible to reconstruct 3-D building models from SAR images.However,in order to obtain sufficient high-resolution imaging capabilities,typical tomographic SAR algorithms usually require dozens of SAR images from different voyages in the same scene.Limited by time and material cost,there are not many SAR images actually available,which makes the traditional tomographic SAR technology unable to produce satisfactory signal-to-noise ratio and height resolution.This paper proposes new solutions and processing methods for how to achieve a more robust,efficient and high-precision tomographic SAR building height inversion under such ill conditions.The main research contents and innovations are as followsIn the first part,a multi-baseline tomographic SAR three-dimensional imaging model was established to study its height imaging principle.The mapping relationship between the three-dimensional structure of the building target and the image features is systematically summarized,and the imaging indicators and imaging conditions are discussed.Aiming at the scattering features such as layover,double bounce,and shadow in SAR images of buildings,the effects of restricted observation conditions on tomographic SAR imaging are analyzed in detail.In the second part,in view of the small number of sailing baselines and the limited length of the synthetic aperture in the height direction,a joint sparse height inversion algorithm for tomographic SAR based on Bayesian information criterion is proposed,which makes full use of the sparseness of SAR image scattering points and buildings The prior information of the target structure is used to optimize the tomographic SAR to reconstruct the distribution of the scatter points in the height direction of the remote sensing target under the condition of limited observation,so as to realize the height inversion of the building.Compared with the traditional orthogonal matching pursuit algorithm,the experimental results show that the algorithm is significantly better than the traditional orthogonal matching pursuit algorithm in both height inversion accuracy and scattering point detection effect.In the third part,a new Conditional Generative Adversarial Network(CGAN)model is designed for the overlapping occlusion of the target scene in the SAR building image.In the construction of the CGAN model,the generator adds a residual network structure,and the input data is directly added to the output layer.Which learns the difference between input and output to solve the problem of network degradation.In addition,the Instance Norm module is introduced to normalize each channel in the convolution layer,which accelerates the network convergence;the discriminator adds a Markov discriminator to enhance the discriminator’s ability to discriminate the authenticity of the data.What’s more,for the problem of insufficient building data in the tomographic SAR scene,the network performs data enhancement on the input slices through random combination,flipping,and translation.Experiments on airborne data show that the reconstruction results using this method are improved in both overlapping and nonoverlapping regions.