Urban 3D imaging using airborne TomoSAR: Contextual information-based approach in the statistical way

Abstract Synthetic aperture radar (SAR) tomography (TomoSAR) technique can eliminate severe overlap in 2D images, and improve target recognition and 3D modeling capabilities, which has become an important trend in SAR development. In recent years, rapid progress has been made and various algorithms were proposed. The Aerospace Information Research Institute, Chinese Academy of Sciences produced the first airborne array TomoSAR system in China. In this paper an airborne TomoSAR experiment using this system and the overall processing flow considering the characteristics of this system are introduced. So far, most of the 3D imaging algorithms used in TomoSAR processing are conducted pixel by pixel, which ignore the structural connections among adjacent pixels. The results will be corrupted by outliers and artifacts due to noise or other uncertainties during processing. To solve this problem, the contextual information of SAR images are explored and utilized by means of the Local Gaussian Markov Random Field (LGMRF) in our 3D reconstruction method. Comparisons with traditional methods are presented to verify the effectiveness of proposed method. The overall processing methods are applied to the data acquired in this experiment and a large scale urban 3D point cloud is obtained, which verifies the ability of the entire system in 3D imaging. This paper presents both the algorithms and experimental results of airborne TomoSAR system, which will contribute in several ways to our understanding of SAR 3D imaging and provide a basis for further research.