Accurate Permittivity Estimation Method for Double-layered Dielectric Object for UWB Radars Takuya Niimi, Shouhei Kidera and Tetsuo Kirimoto Graduate School of Informatics and Engineering, University of Electro-Communications, Tokyo, Japan Abstract—A microwave ultra-wideband (UWB) radar system is one of the most promising technologies for the internal imaging of dielectric objects in applications such as non-invasive inspec- tion and non-destructive measurement. For such applications, we have already proposed an efficient method of estimating permittivity employing a geometrical optics (GO) approximation and exploiting dielectric outer boundary points and their normal vector accurately obtained with the range points migration (RPM) method. However, this method assumes that there is a highly lossy object buried in a single-layered homogeneous dielectric medium. This paper proposes an accurate permittivity estimation method for a double layered dielectric medium that employs the RPM or Envelope method extended to boundary extraction for each dielectric layer, and an estimation of time delay based on a geometrical optics approximation. The results from finite-difference time-domain (FDTD) based simulations demonstrate that our method accurately estimates permittivity and the boundary of the two layers simultaneously. Key words—UWB sensor, Range points migration (RPM), Envelope method, Inverse scattering problem. I. INTRODUCTION There are emerging demands for an innovative technique of imaging objects embedded in dielectric media. Such a technique would be used in non-invasive screening of the human body for medical purposes and the monitoring of artificial structures to reduce damage in the event of a natural catastrophe. One of the most promising technologies is UWB radar system that has sufficient range resolution and the ability to penetrate dielectric objects, especially at lower microwave frequencies. For such application, there are various inverse scattering methods for reconstructing the dielectric constant, such as that based on domain integral equations [1]. How- ever, the size of space discretization must be severely con- strained to avoid sluggish convergence in multi-dimensional optimizations. While methods such as those in [2] require less computational resources owing to their GO approximation, another approach assumes a simple and known structure of the dielectric medium, such as a cuboid. We have already developed a promising method of simul- taneously obtaining an accurate internal image and estimating permittivity [3]. This method employs the original method of range points migration (RPM) [4] to correctly reconstruct dielectric boundary points and their normal vectors. The actual time delay of the microwave propagating through the dielectric medium can then be accurately estimated from the recorded transmissive data. However, this method is specific to a simple target model in which highly lossy material is buried in a single-layered homogeneous medium. Fig. 1. System model. To remove this limitation, this paper proposes a novel method of accurately estimation permittivity and extracting the boundary within a double-layered medium. In this method, the extended Envelope method is introduced to reconstruct the inner boundary points and their normal vectors, and GO based permittivity estimation is adopted to reduce the required computational resources. The results of numerical simulation show that our proposed method simultaneously estimates per- mittivity accurately and extracts the boundary within the inner medium with accuracy on the order of 1/20 of the wavelength of the transmitting signal. II. SYSTEM MODEL Figure 1 shows the system model. It is assumed that a dielectric object has a double layered structure. The regions of the outer and inner media are denoted Ω 1 and Ω 2 , respectively. Two omni-directional antennas scan along a circle with center  c and radius  c that completely surrounds the dielectric object. One transmitting and receiving antenna is located at  TR =(,  ), and an antenna playing only a receiving role is located at  R =(,  ), where  c =( TR +  R )/2 holds. A mono-cycle pulse with center wavelength  is used as the transmitting current.  TR (, , ) and  R (, , ) denote the output of the matched filter at antenna positions  TR and  R , respectively, where  =c/2 is expressed by time  and the propagation speed of the radio wave c. The range points extracted from the local maxima of  TR and  R are debited  TR =( TR , TR , TR ) and  R =( R , R , R ), respectively, and the detailed process of the extraction has been described previously [4]. The sets of these range points  TR and  R are respectively denoted  TR and  R . 95 WE3A_02 WE3A_02 WE3A_02 WE3A_02 Proceedings of ISAP 2014, Kaohsiung, Taiwan, Dec. 2-5, 2014