Polarization-Aware Ray-Tracing Enhanced Back-Projection Algorithm for Microwave Imaging in Complex Multipath Environments
Han Na \affref ref1, Quanfeng Wang \affref ref1, Matthias Saurer \affref ref1, Meisong Tong \affref ref2, and Thomas F. Eibert \affref ref1 · 2026
Read Polarization-Aware Ray-Tracing Enhanced Back-Projection Algorithm for Microwave Imaging in Complex Multipath Environments by Han Na \affref ref1, Quanfeng Wang \affref ref1, Matthias Saurer \affref ref1, Meisong Tong \affref ref2, and Thomas F. Eibert \affref ref1 (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-10081.txt.
Introduction
Na and four colleagues improve microwave imaging in multipath settings, where signals bounce off surroundings before reaching the target. Their ray-tracing enhanced back-projection algorithm weaves ray tracing into a generalised back-projection method, fixing two weaknesses of the standard approach. First, reflected paths can image hidden targets when the direct line of sight is blocked.
Second, using reflected paths acts like widening the aperture, sharpening resolution with no new measurements. The key enabler is treating waves as vectors, with polarisation-dependent phase compensation that scalar models ignore. A shooting-and-bouncing-rays framework adds flexibility over manual path evaluation in earlier ray-tracing back-projection methods, yielding cleaner images in complex scenes.
Catalog record
| CREATOR | Han Na \affref ref1, Quanfeng Wang \affref ref1, Matthias Saurer \affref ref1, Meisong Tong \affref ref2, and Thomas F. Eibert \affref ref1 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 16 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.10081
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Polarization-Aware Ray-Tracing Enhanced Back-Projection Algorithm for Microwave Imaging in Complex Multipath Environments Han Na \affref ref1, Quanfeng Wang \affref ref1, Matthias Saurer \affref ref1, Meisong Tong \affref ref2, and Thomas F. Eibert \affref ref1 Abstract A ray-tracing (RT) enhanced back-projection algorithm (RT-BPA) for microwave imaging in multipath environments is presented. By tightly incorporating the concept of ray-tracing into a generalized version of traditional BPA, this method ensures improved image quality by addressing two important issues. First, when the line-of-sight (LOS) path is obstructed, reflected paths, if available, enable imaging of hidden targets, which extends the applicability of the standard BPA beyond its normal use case. Second, the consideration of reflected ray-paths is equivalent to virtually increasing the aperture size, thus, improving image resolution without requiring new measurements. A key factor in achieving these advancements is the consideration of the vector nature of electromagnetic waves with polarization-dependent phase compensation, which is often ignored when employing a scalar-wave based formulation of the electromagnetic vector field. In addition, the presented method employs a shooting and bouncing rays (SBR) framework, offering better flexibility compared to manual path evaluation in existing RT-BPAs. 1 Introduction In microwave imaging, multipath effects caused by environmental scattering can be both beneficial and challenging. According to existing literature, these effects can be viewed from two perspectives. One possibility is to treat reflections from objects other than the target of interest as noise or clutter that requires suppression [ 1 ] . Alternatively, such reflections can be interpreted and utilized as indirect viewing mechanisms, enabling imaging when the direct line-of-sight (LOS) path is blocked [ 2 ] or improving image resolution by effectively increasing the synthetic aperture [ 3 ] . In this work, we focus on the latter perspective and introduce a ray-tracing (RT) enhanced back-projection algorithm (RT-BPA) for microwave imaging in multipath environments. To address the challenges posed by multipath propagations, various studies have employed ray-based techniques to identify multipath contributions and correct signal phase based on ray-path lengths [ 4 , 5 ] . However, most of these approaches focus on the signal processing perspective, neglecting critical physical properties such as the polarization of electromagnetic waves. Without proper consideration of polarization effects, the coherent combination of various signal contributions from multiple paths becomes inaccurate. Thus, the quality of reconstructed images is reduced. Furthermore, many of these methods rely on fixed or simplified environments, explicitly calculating ray-paths for specific cases [ 4 ] and often restricting the analysis to 2-D scenarios [ 2 ] . Advanced imaging methods based on solving the inverse electromagnetic problem can theoretically address these limitations, but their reliance on analytically known Green’s functions [ 6 , 7 ] makes them impractical for non-canonical or complex propagation environments. This work introduces an enhancement to the standard formulation of the naive BPA using a general RT engine based on the CUDA OptiX framework [ 8 ] . By employing the shooting and bouncing rays (SBR) method, the algorithm dynamically identifies ray-paths, enabling simulations of complex 3-D environments without manual path calculations. Employing a polarization-aware description of the scattering process, the proposed method improves directional resolution and serves as a valuable tool for creating well-focused microwave images in complex propagation environments. 2 Theoretical Background Microwave imaging algorithms, including BPA, are commonly based on a linearization of the forward scattering problem according to the first-order Born [ 6 , 9 ] , which neglects mutual coupling between different points of the target scatterer. Due to this linearization, the forward scattering operator is expressed as an integral operator connecting the dyadic scattering distribution \boldsymbol{\bar{s}}_{\mathrm{B}}(\boldsymbol{r}) within the volume V_{\mathrm{S}} to the measurement samples T(\boldsymbol{r}_{\mathrm{T}},\boldsymbol{r}_{\mathrm{R}},k) according to [ 9 ] T(\boldsymbol{r}_{\mathrm{T}},\boldsymbol{r}_{\mathrm{R}},k)=\iiint\limits_{V_{\mathrm{S}}}\boldsymbol{E}_{\mathrm{R}}\left({\boldsymbol{r}},\boldsymbol{r}_{\mathrm{R}}\right)\cdot\boldsymbol{\bar{s}}_{\mathrm{B}}\left({\boldsymbol{r}}\right)\cdot\boldsymbol{E}_{\mathrm{T}}\left({\boldsymbol{r},\boldsymbol{r}_{\mathrm{T}}}\right)\,\mathrm{d}^{3}\boldsymbol{r}\,, (1) where k is the wavenumber while \boldsymbol{r}_{\mathrm{T}} and \boldsymbol{r}_{\mathrm{R}} are the position vectors of the transmitter (Tx) and receiver (Rx), respectively. Additionally, the expression \boldsymbol{E}_{\mathrm{T}}\left({\boldsymbol{r},\boldsymbol{r}_{\mathrm{T}}}\right) in ( 1 ) characterizes the electric field vector of the Tx antenna, which is incident on the scattering object, while \boldsymbol{E}_{\mathrm{R}}\left({\boldsymbol{r},\boldsymbol{r}_{\mathrm{R}}}\right) is the scattered field vector, which is reciprocal to the field radiated by the Rx antenna, when it is used in transmit mode [ 9 ] . In the standard BPA, the wave propagation environment between each Tx and the scatterer, as well as between the scatterer and the Rx, is assumed to be free space [ 10 ] . Denoting polarization vectors \boldsymbol{\hat{E}}_{\mathrm{T}}(\boldsymbol{r},\boldsymbol{r}_{\mathrm{T}}) and \boldsymbol{\hat{E}}_{\mathrm{R}}(\boldsymbol{r},\boldsymbol{r}_{\mathrm{R}}) , respectively, and decomposing the field propagation operator into a linear phase shift term \exp\left[-\mathrm{j}\varphi(\boldsymbol{r})\right] and a magnitude term A(\boldsymbol{r}_{\mathrm{T}},\boldsymbol{r}_{\mathrm{R}},\boldsymbol{r}) 1 1 1 In this work, the attenuation factors are set to 1 for arbitrary \boldsymbol{r}_{\mathrm{T}} , \boldsymbol{r}_{\mathrm{R}} and \boldsymbol{r} by employing the paraxial approximation. , ( 1 ) can be rewritten as T(\boldsymbol{r}_{\mathrm{T}},\boldsymbol{r}_{\mathrm{R}},k)=\iiint\limits_{V_{\mathrm{S}}}\boldsymbol{\hat{E}}_{\mathrm{R}}(\boldsymbol{r},\boldsymbol{r}_{\mathrm{R}})\cdot\boldsymbol{\bar{s}}_{\mathrm{B}}\left({\boldsymbol{r}}\right)\cdot\boldsymbol{\hat{E}}_{\mathrm{T}}(\boldsymbol{r},\boldsymbol{r}_{\mathrm{T}})\\ A(\boldsymbol{r}_{\mathrm{T}},\boldsymbol{r}_{\mathrm{R}},\boldsymbol{r})\mathrm{e}^{-\mathrm{j}\varphi(\boldsymbol{r})}\mathrm{d}^{3}\boldsymbol{r}\,, (2) where \varphi(\boldsymbol{r})=k\left[R_{\text{Rx}}(\boldsymbol{r})+R_{\text{Tx}}(\boldsymbol{r})\right] with R_{\text{Tx}}(\boldsymbol{r})=\left|\boldsymbol{r}-\boldsymbol{r}_{\mathrm{T}}\right| and R_{\text{Rx}}(\boldsymbol{r})=\left|\boldsymbol{r}-\boldsymbol{r}_{\mathrm{R}}\right| denoting the corresponding path lengths between the antennas and a specific point located in the imaging domain. However, this formulation neglects the wave interaction with the background objects, e.g., walls and ceilings in an indoor scenario [ 1 ] . With the consideration of reflective propagation environments, we separate the field vectors interacting with the scattering object into a superposition of distinct wavefronts characterized by a unique index w , and ( 2 ) is generalized to T(\boldsymbol{r}_{\mathrm{T}},\boldsymbol{r}_{\mathrm{R}},k)=\iiint\limits_{V_{\mathrm{S}}}\sum\limits_{w}\boldsymbol{\hat{E}}^{w}_{\mathrm{R}}(\boldsymbol{r},\boldsymbol{r}_{\mathrm{R}})\cdot\boldsymbol{\bar{s}}_{\mathrm{B}}\left({\boldsymbol{r}}\right)\cdot\boldsymbol{\hat{E}}^{w}_{\mathrm{T}}(\boldsymbol{r},\boldsymbol{r}_{\mathrm{T}})\\ A^{w}(\boldsymbol{r}_{\mathrm{T}},\boldsymbol{r}_{\mathrm{R}},\boldsymbol{r})\mathrm{e}^{-\mathrm{j}\varphi^{w}(\boldsymbol{r})}\mathrm{d}^{3}\boldsymbol{r}\,. (3) By taking the adjoint of the generalized forward scattering operator in multipath environments, as given in ( 3 ), and considering only one set of co-polarized scattering data for a fair comparison with the conventional BPA, we get s_{\mathrm{B}}(\boldsymbol{r})=\sum\limits_{w}\int\limits_{0}^{\infty}\iint\limits_{L_{\mathrm{T}}}\iint\limits_{L_{\mathrm{R}}}A^{w}(\boldsymbol{r}_{\mathrm{T}},\boldsymbol{r}_{\mathrm{R}},\boldsymbol{r})T(\boldsymbol{r}_{\mathrm{T}},\boldsymbol{r}_{\mathrm{R}},k)\\ \mathrm{e}^{\,\mathrm{j}\varphi^{w}(\boldsymbol{r})+\mathrm{j}\uppi\delta(w)}\mathrm{d}^{2}\boldsymbol{r}_{\mathrm{R}}\,\mathrm{d}^{2}\boldsymbol{r}_{\mathrm{T}}\,\mathrm{d}k\,, (4) where L_{\mathrm{T}} and L_{\mathrm{R}} represent the aperture of the transmit and receive array, respectively. The term \delta(w) is either 1 for parallel polarization or 0 for vertical polarization. This formulation accounts for a \uppi -phase shift due to potential polarization changes, a phenomenon also known as the half-wave loss in optics [ 11 ] . By comparison, the reconstruction formula for the naive BPA is given by s_{\mathrm{B}}(\boldsymbol{r}^{\prime})=\int\limits_{0}^{\infty}\iint\limits_{L_{\mathrm{T}}}\iint\limits_{L_{\mathrm{R}}}T(\boldsymbol{r}_{\mathrm{T}},\boldsymbol{r}_{\mathrm{R}},k)\,\mathrm{e}^{\,\mathrm{j}\varphi(\boldsymbol{r})}\mathrm{d}^{2}\boldsymbol{r}_{\mathrm{R}}\,\mathrm{d}^{2}\boldsymbol{r}_{\mathrm{T}}\,\mathrm{d}k, (5) which does not account for multipath effects. By incorporating the separation of different wavefronts and a polarization-specific phase-compensation, the RT-BPA, expressed in ( 4 ), is expected to achieve improved focusing capabilities compared to the naive BPA in ( 5 ). With slight modifications, e.g., dropping the integral over the receive array, ( 4 ) and ( 5 ) can also be utilized in radiation problems. With the proper consideration of different wavefronts, the generalized BPA expressed by ( 4 ) can be easily incorporated with ray-tracing frameworks using the SBR approach. Here, rays are randomly launched from each image voxel and propagate through a 3-D environment modeled by a triangular mesh following geometrical optics (GO) principles. Each ray captured by a Tx or an Rx provides the optical path length information and determines whether an extra \uppi -phase shift is introduced in ( 4 ). The latter is managed by assigning an integer hash identifier w to each individual path, which depends on the sequence of the environmental objects it interacts with. Since all rays experiencing the same bounces share a common identifier, w , the post-processing of the extra phase shift denoted by \delta(w) in ( 4 ) is easy and efficient. The final image is then created by coherently summing up all signal contributions from the obtained wavefronts. To manage the