Simulation of complex target RCS with application to SAR image recognition

• Main Authors: Cheng-Yen Chiang, Kun-Shan Chen
• Format: Conference Proceeding
• Language: English
• Subjects: Accuracy; Atmospheric modeling; Computational modeling; Radar cross section; Solid modeling; Synthetic aperture radar; Target recognition

SAR(synthetic aperture radar) image understanding and interpretation is essential for remote sensing of earth environment and target detection. In development of aided target recognition and identification system, SAR image database with rich information content plays important role is essential. This paper presents a RCS computation for simulation of orbital SAR image. After demodulation, the received SAR signal is given as [1-2] s 0 (τ, η) = A 0 w r (τ - (2R(η))/c, T p )w a (η) exp{- j (4πf c R(η))/c + jπα r (τ - (2R(η))/c) 2 + φ"} (1) where φ" is lumped sum of phase noise from atmosphere, satellite altitude error, terrain, etc. and R is distance from antenna to target being observed, A 0 is slant range backscatter coefficient of the target, φ the phase term, and w a (η) is antenna pattern and is a function of slow time. To take into account the radar backscattering characteristics in (5), we apply Radar Cross Section Analysis and Visualization System (RAVIS) [3] that utilizes the physical optics (PO), physical diffraction theory (PDT), and shooting and bouncing rays (SBR) to compute the RCS of complex radar targets [4-8]. Single scattering and diffraction from a target are first computed by PO and PDT, followed by SBR to account for multiple scattering and diffraction. The system outputs for a given 3D CAD model of the target of interest. The CAD model contains numerous grids or polygons, each associated with computed RCS as function of incident and aspect angles for a given set of radar parameters. The number of polygons is determined by target's geometry complexity and its electromagnetic size. To realize the imaging scenario, each polygon must be properly oriented and positioned based on proper coordinates system. SAR image is sensitive to target's geometry including orientation and aspects angles. For target recognition and identification, more complete database for feature extraction is preferable to achieve better performance and reduce false alarm rate. In SAR image simulation, supposed N samples (incident angles and aspect angles) is desired, then the computation complexity is O(N 3 ). Statistics indicate that with 1.97 GB RAM and 2.4GHz (4 cores) CPU to complete one TerraSAR-X image simulation of MD80 aircraft using 25672 polygons representing RCS for a pose (one incident angle and one aspect angle) it would take about 11 months to complete all poses for incident angle from 20° -50° with 5° a step and aspect angles from -180° ~+180° with 1° a s