We consider a multistatic radio-frequency imaging problem with anisotropy, in which the reflection from a point depends on the positions of the transmit (Tx) and receive (Rx) arrays. The goal is to label the voxels of a field of view by a finite set of semantic classes and to group them into object instances. For the image formation of each Tx--Rx pair we apply a standard inverse-problem solver, and we feed the resulting per-pair reconstructions into a trained three-dimensional (3-D) U-Net that performs the fusion implicitly and the per-voxel classification explicitly. On a controlled, under-determined multistatic setup, we consider the following image formation methods: back-projection (BP) and the least absolute shrinkage and selection operator (LASSO) from a single deterministic snapshot, and incoherent BP and group-LASSO from multiple fading snapshots. For each imaging method we train a separate U-Net that fuses the six Tx--Rx pairs (its input channels) and assigns each voxel a probability vector over the classes. Taking the most probable class gives a labeled volume---the semantic reconstruction. Object instances and their oriented bounding boxes then follow by geometric post-processing (clustering and principal-component analysis). Across a wide range of signal-to-noise ratio, the semantic reconstruction (scored against ground truth by segmentation intersection-over-union) and the resulting 3-D detection degrade far more gracefully than the classical intensity reconstruction: the detection in particular stays reliable well into noise levels at which that reconstruction has dissolved. Because real scenes contain objects of classes the network was not trained on, we add an explicit unknown class trained by outlier exposure, which labels held-out novel objects as unknown instead of mislabeling them as a known class by reconstructed shape.
Figures & tables
Fig. 1: End-to-end pipeline on one example scene (schematic). Left: the ground-truth (GT) scene in the field of view (FoV), observed by three uniform planar array (UPA) terminals A, B, C outside the FoV; the six Tx–Rx pairs are the coloured arrows, and the per-pair reconstructions c^i,j form the six input channels. Top branch (conventional): their arithmetic mean is the naive intensity fusion Iˉ , with no class labels. Bottom branch (learned): a 3-D U-Net fuses the six channels and outputs, per voxel, a probability vector pc(q) over the six classes. Right (post-processing, no learning): the voxels of each foreground class are clustered (DBSCAN) and fitted with one oriented bounding box (PCA); combining them gives the fused semantic reconstruction with detected boxes (top view), contrasted (“vs”) with the naive fusion.
Fig. 2: One open-set scene (known human/tree/bench and held-out novel objects), viewed from between the panels of pair (1,2) . Left: the per-pair coupling ε1,2,q (Eq. ( 2 )) on the scene surface—bright where the pair sees the surface strongly—with its bistatic ray paths. Right: the learned semantic reconstruction (group-LASSO, ρ=−20 dB), voxels coloured by predicted class, novel objects flagged unknown (magenta).
Fig. 3: Point-spread function on a single canonical sphere carrying the per-pair anisotropy ε ( xy cut at z=1 m, dB relative to peak). Columns are the four imaging methods, rows the noise level ρ=−25 and −45 dB. Each panel is the naive intensity fusion (not the U-Net), the sphere footprint marked by the white circle.
Fig. 4: Reconstruction quality of the naive intensity fusion (Eq. ( 8 )) versus ρ on the test set: the threshold-free Esdf and the Chamfer distance (both in metres, lower is better; ρ increases left to right toward clean).
Fig. 5: Naive intensity fusion versus the learned semantic reconstruction on one open-set test scene. Rows are the four imaging methods, columns the naive and learned reconstructions at ρ=−25 and −45 dB, and each boxed pair shares a method and ρ . Naive: top-down maximum-intensity projection (dB). Learned: the top-down class-labeled map ℓ^ (known classes coloured, novel objects flagged unknown in magenta).
Fig. 6: Learned-fusion performance versus ρ (one U-Net per imaging method, test set): mIoU, known recall, object-level open-set AUROC, and the fraction of held-out novel objects mislabeled as a known class (the bar panel “novel → known”, at clean SNR).
Reconstructing object geometry from radio frequency (RF) signals is fundamentally challenging due to the lensless imaging nature of RF sensing, which leads to low spatial resolution and high noise. Unlike light signals, RF signals can penetrate occlusions and thus capture information about hidden scenes. Existing Non-Line-of-Sight (NLoS) 3D neural reconstruction methods can recover coarse surfaces inside enclosed environments but often suffer from unstable optimization, noisy surface geometry, and surface ambiguity, failing to produce accurate zero-level sets from the signed distance field (SDF). These limitations largely stem from neglecting the role of Line-of-Sight (LoS) geometry outside the enclosed region, which provides valuable physical constraints for modeling signal propagation. In this paper, we introduce a Unified LoS and NLoS neural geometry reconstruction framework GeRaF 2.0 that leverages the outside LoS geometry to model and guide RF propagation from the LoS region into the NLoS region. By integrating visual LoS priors into the neural field formulation, GeRaF 2.0 achieves stable training and physically consistent reconstruction of both visible and hidden geometry, setting a new state-of-the-art in RF-based geometry reconstruction.
GeRaF is the first method to use neural implicit learning for near-range 3D geometry reconstruction from radio frequency (RF) signals. Unlike RGB or LiDAR-based methods, RF sensing can see through occlusion but suffers from low resolution and noise due to its lensless imaging nature. While lenses in RGB imaging constrain sampling to 1D rays, RF signals propagate through the entire space, introducing significant noise and leading to cubic complexity in volumetric rendering. Moreover, RF signals interact with surfaces via specular reflections, requiring fundamentally different modeling. To address these challenges, GeRaF (1) introduces filter-based rendering to suppress irrelevant signals, (2) implements a physics-based RF volumetric rendering pipeline, and (3) proposes a novel lensless sampling and lensless alpha blending strategy that makes full-space sampling feasible during training. By learning signed distance functions, reflectiveness, and signal power through MLPs and trainable parameters, GeRaF takes the first step towards reconstructing millimeter-level geometry from RF signals in real-world settings.
Ultrasound imaging increasingly targets portable, point-of-care, and wearable settings where constraints on power, bandwidth, and hardware complexity often necessitate sparse data acquisition in spatiotemporal scanning. However, image reconstruction using the sparse data can introduce insufficient phase information in coherent beamforming process, resulting in grating-lobe artifacts that degrade imaging contrast resolution. We present a physics-guided, data-driven framework for sparse-to-dense radio-frequency (RF) reconstruction that aligns training with downstream image formation. Our approach trains an end-to-end interpolation network using a hybrid supervision scheme that combines an RF-domain and a beamforming-domain loss with exponential moving average (EMA) to stabilize the multi-objective training. To improve generalization under variable acquisition layouts, we also introduce a random-skip masking strategy that varies sparsity patterns during training so a single model can handle diverse decimation factors and irregular channel configurations. We evaluate the framework on a held-out test set using the mean structural similarity index measure (SSIM) between reconstructed and ground-truth beamformed images. Across decimation factors ×2 to ×13, the best-performing configuration maintains mean SSIM around 0.95. Overall, the results show consistent gains in RF reconstruction and post-beamforming image quality across diverse acquisition conditions. This approach enables robust, high-quality ultrasound imaging at resource-constrained settings by allowing more sparse scanning in spatiotemporal domain.
Luoyuan Zhang, Yiyang You, Ananya Tandri +4
University of North Carolina at Chapel Hill, Chapel Hill, NC, USA · Johns Hopkins University, Baltimore, MD, USA