TOMOGRAPHIC RECONSTRUCTIONS OF BUILDINGS IN PARIS The 3D reconstruction results may be obtained with various methods (Figures S6-S11). (a) (Fast-Sup-GLRT Parameters Used: KM = 2, PFA = 10-3, T1 and T2 Computed Through Monte-Carlo Simulation) (b) FIGURE S6. (a) The multiple signal classification estimator is followed by (b) postprocessing (geometrical transformation from radar coordinates to ground coordinates). (a) -5 (m) 35 100 m (a) (b) FIGURE S7. (a) Compressed sensing is followed by (b) postpro- cessing (geometrical transformation from radar coordinates to ground coordinates) with one scatter by pixel. (b) (c) FIGURE S9. The Generalized Likelihood Ratio Test (GLRT)- FIGURE S8. Reflectivity is obtained through spatially regularized 3D inversion. based detection approach. The Fast-Support (Fast-Sup)-GLRT parameters used were KM = 2, PFA = 10 -3. T1 and T2 were computed through a Monte Carlo simulation. (a) A 3D point cloud, color-coded by height. (b) The detected double scatters. Red indicates lowest, and blue indicates highest. (c) The surface reconstructed from the point cloud. (Continued ) JUNE 2020 IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE 17