Determination of height of urban buildings based on non-parametric estimation of signal spectrum in SAR data tomography

Nowadays, the TomoSAR technique has been able to overcome the limitations of radar interferometry techniques in separating multiple scatterers of pixels. By extending the principles of virtual aperture in the elevation direction, these techniques pay much attention in the analysis of urban challenging areas. Despite the expectation of interference of the distribution of buildings with different heights, wall and ceiling levels, ground, or tree trunks, the TomoSAR technique enables the retrieval of reflectivity function along elevation direction by applying a stack of images taken at different times and slightly different orbit positions. In this research, the capability of a new non-parametric method of signal spectrum estimation in the analysis of the third dimension of buildings in urban areas is investigated. The proposed efficient maximum entropy estimator has been able to provide good performance in reconstructing three-dimensional information from the urban area. By searching for coefficients of the automatic regression model, this method can maximize the signal entropy, also separate the different noise levels. Implementation of the proposed algorithm using 19 TerraSAR-X satellite images in the study of the third-millennium tower in Tehran with a height of 120 meters and comparison of the results with capon and Beamforming spectral estimation methods, indicates better performance of the proposed method in height reconstruction, the continuity reflectivity profile and the elimination of the effects of the side lobes.