Coherent Signal DOA Estimation and 3D Point Cloud Imaging Based on a Straight-Curved Hybrid L-Shaped Conformal Array

Abstract

Highlights What are the main findings? A straight-curved hybrid L-shaped conformal array topology tailored for cylindrical radomes is proposed, with an orthogonal virtual manifold transformation that compensates for non-linear phase distortion induced by the conformal surface. A cascaded Forward-Backward Spatial Smoothing (FBSS) and Root-MUSIC framework is developed for efficient coherent signal decoherence, with a Frobenius-norm-based global cost function that pairs independent angles and suppresses spatial ghost targets. What are the implications of the main findings? Systematic 3D computer vision evaluations demonstrate a median point-wise localization error of 0.3986 m, CD of 0.9695 m2, EMD of 1.9504 m, and an angular resolution of 1.0° under the stated simulation conditions; the algorithm maintains stable 3D reconstruction around a post-pulse-compression SNR of −8.0 dB and a simulation-based curvature threshold of 12.50 m−1 (r=0.08 m). This framework provides a data processing pipeline for high-resolution conformal radar 3D perception and offers theoretical support and dimensional design references for next-generation airborne radar system design. Abstract High-speed airborne platforms impose stringent aerodynamic constraints that restrict traditional planar antenna designs. Concurrently, multi-component echoes from extended targets and complex propagation environments induce strong signal coherence, severely degrading spatial angle estimation. To address these dual challenges, this paper proposes a straight-curved hybrid L-shaped asymmetric conformal array hardware topology tailored for cylindrical radomes. Building upon this, a millimeter-wave radar 3D point cloud imaging framework is developed for coherent targets. An orthogonal virtual manifold transformation is first devised to effectively compensate for the non-linear phase distortion induced by the conformal topology. Subsequently, a cascaded Forward-Backward Spatial Smoothing (FBSS) and Root-MUSIC framework is used for efficient signal decoherence. To resolve angle mismatches, a global cost function based on the cross-covariance Frobenius norm is formulated, which pairs independent angles and significantly suppresses spatial ghost targets. Systematic evaluations using 3D computer vision metrics demonstrate that the proposed method achieves accurate geometric restoration of aircraft targets with coherent signals. In the representative simulation, the method obtains a median point-wise localization error of 0.3986 m, a Chamfer Distance (CD) of 0.9695 m2, an Earth Mover’s Distance (EMD) of 1.9504 m, and a spatial angular resolution of 1.0° under the stated test conditions. Under the stated simulation assumptions, boundary analyses indicate stable reconstruction around a post-pulse-compression SNR of −8.0 dB and a conformal curvature of 12.50 m−1 (r=0.08 m), providing simulation-based design references for conformal radar 3D imaging.