Covariance-based Imaging and Multi-View Fusion for Networked Sensing

TL;DR

This work tackles imaging of extended targets in 6G ISAC networked sensing by proposing a two-phase pipeline. Phase I performs covariance-based single-view imaging at each receive BS by estimating effective scattering intensities ${\gamma_{r,k}}$ and dynamic grid positions ${\mathbf{P}_k}$ from the sample covariance ${\hat{\bm{\Sigma}}_k}$, reducing parameter dimensionality versus instantaneous-channel methods. Phase II aligns and fuses these heterogeneous single-view maps using edge-preserving natural neighbor interpolation (EP-NNI) and a joint optimization over fused scattering ${\bm{\gamma}'_r}$ and receiver informativity ${\bm{\Lambda}}$, incorporating sparsity and total-variation regularization via ADMM. The approach leverages target geometry and FoV diversity to outperform FFT/CS-based and prior fusion schemes, achieving sharper boundaries, lower sidelobes, and higher IoU under resource constraints. Overall, the framework enables high-quality environment reconstruction for future 6G networked sensing applications by exploiting phase-coherent covariance structure and cross-view information with geometry-aware fusion.

Abstract

This paper considers multi-view imaging in a sixth-generation (6G) integrated sensing and communication network, which consists of a transmit base-station (BS), multiple receive BSs connected to a central processing unit (CPU), and multiple extended targets. Our goal is to devise an effective multi-view imaging technique that can jointly leverage the targets' echo signals at all the receive BSs to precisely construct the image of these targets. To achieve this goal, we propose a two-phase approach. In Phase I, each receive BS recovers an individual image based on the sample covariance matrix of its received signals. Specifically, we propose a novel covariance-based imaging framework to jointly estimate effective scattering intensity and grid positions, which reduces the number of estimated parameters leveraging channel statistical properties and allows grid adjustment to conform to target geometry. In Phase II, the CPU fuses the individual images of all the receivers to construct a high-quality image of all the targets. Specifically, we design edge-preserving natural neighbor interpolation (EP-NNI) to map individual heterogeneous images onto common and finer grids, and then propose a joint optimization framework to estimate fused scattering intensity and BS fields of view. Extensive numerical results show that the proposed scheme significantly enhances imaging performance, facilitating high-quality environment reconstruction for future 6G networks.

Figures (5)

• Figure 1:
• Figure 2: Covariance-based single-view images at receiver $1$ with $N_{\rm tx} = 16, N_{\rm rx} = 16, L = 16, M = 20, P = 10$ dBm, and $Q = 400$: (a) Fixed grid without penalty \ref{['eq:distributed_penalty']}; (b) Fixed grid with penalty \ref{['eq:distributed_penalty']}; (c) Optimized grid without penalty \ref{['eq:distributed_penalty']}; (d) Optimized grid with penalty \ref{['eq:distributed_penalty']}; (e) NNI; (f) EP-NNI.
• Figure 3: Covariance-based single-view images at receiver $2$ with $N_{\rm tx} = 8, N_{\rm rx} = 8, M = 20, L = 8$, and $P = 10$ dBm: (a) Grid optimization with $Q = 900$ and interpolation to uniform grids with $Q' = 60\times 60$; (b) Fixed uniform grids with $Q = 60\times 60$; (c) Fixed uniform grids with $Q = 90\times 90$; (d) Fixed uniform grids with $Q = 120\times 120$.
• Figure 4: Covariance-based single-view images and multi-view fusion results with $N_{\rm tx} = N_{\rm rx} = L = 8, M = 20, P = 0$ dBm, $Q = 900$, and $Q' = 60 \times 60$: (a) Image at receiver $1$; (b) Image at receiver $2$; (c) Image at receiver $3$; (d) Multi-view fusion with WLS penalty; (e) Multi-view fusion with WLS and sparse penalties; (f) Multi-view fusion with WLS, sparse, and TV penalties.
• Figure 5: Imaging performance comparison between different schemes with $K = 3, Q = 900$, and $Q' = 60 \times 60$. The first, the second, and the third rows present the images of benchmark scheme I, benchmark scheme II, and our proposed scheme, respectively. The first column is obtained assuming $N_{\rm tx} = N_{\rm rx} = L = 24, M = 20$, and $P = 10$ dBm. The second column is obtained assuming $N_{\rm tx} = N_{\rm rx} = L = 12, M = 20$, and $P = 10$ dBm. The third column is obtained assuming $N_{\rm tx} = N_{\rm rx} = 24, L = 12, M = 20$, and $P = 10$ dBm. The fourth column is obtained assuming $N_{\rm tx} = N_{\rm rx} = L = 24, M = 20$, and $P = -10$ dBm. The fifth column is obtained assuming $N_{\rm tx} = N_{\rm rx} = L = 24, M = 5$, and $P = 10$ dBm.