A Multi-Scale Global Denoising Algorithm for Urban Airborne Array InSAR Point Clouds

Array Interferometric Synthetic Aperture Radar (Array InSAR) enables high-precision 3D point cloud reconstruction, but data quality is often degraded by environmental noise, channel amplitude–phase errors, and multipath interference. To address the main noise types in urban airborne Array InSAR point clouds—outlier noise, channel amplitude-phase errors, and multipath interference—this paper proposes a multi-scale global denoising algorithm tailored to their characteristics. The proposed method first removes prominent airborne outlier noise using a denoising algorithm based on an adaptive dispersion coefficient. It then adjusts point cloud normal based on cosine similarity and information entropy. To mitigate channel amplitude-phase noise, an adaptive weighted filter incorporating curvature and intensity features is proposed. Finally, multipath interference is mitigated through geometric and backscattering intensity consistency verification. Experimental results demonstrate that MSGDA achieves an ONRI of 0.9184, an EPI above 0.8, a CCI of 0.8134, and a DUI of 0.7307, consistently outperforming comparison algorithms. These results confirm that MSGDA effectively suppresses noise while preserving structural features, thereby enhancing the quality and reliability of Array InSAR point clouds for urban applications.