Talk · International Siberian workshops and tutorials on electron devices and materials · June 1, 2026
Robustness of Spectral-Geometric Invariants Under Realistic 3D Acquisition Distortions: A Critical Review, Taxonomy, and Comparative Evaluation Protocol
Raxmanov Raxmanbergan Abdullaevich, Худайберганов Темур Рустамович, Bozorboy Palvanov, Temur Takhirovich Turdiyev, Javlonbek Uralov
Abstract
-spectral-geometric invariants play a key role in 3D shape analysis because they encode intrinsic structure through differential operators and are largely insensitive to rigid motion. However, real 3D acquisition and modeling pipelines (photogrammetry, LiDAR, structured light, and medical reconstruction) seldom produce ideal watertight meshes; instead, they contain irregular sampling, anisotropic remeshing, noise, holes, partial scans, and topology defects that perturb discrete operators and reduce descriptor stability. This paper presents a distortion-aware review of spectral–geometric invariants for retrieval, classification, and shape comparison. We provide three contributions: (1) a taxonomy organized by generating operator (Laplace–Beltrami spectra, heat/diffusion kernels, wave kernels), descriptor scope (global, pointwise, functional/operator), and invariance target (rigid, scale, nearisometric, and partial robustness); (2) a unified evaluation protocol, the Robustness Envelope, which quantifies descriptor drift, discriminative-margin loss, and task-level accuracy degradation as functions of distortion severity; and (3) a synthesis of failure modes with deployment-oriented reporting and implementation recommendations. Evidence across distortion families indicates that diffusion signatures usually offer the best robustness–discriminativity tradeoff under moderate noise and resolution changes; eigenvalue fingerprints remain compact but are fragile to domain change; and wave/functional approaches are strong when eigenbasis quality and correspondence constraints are reliable.
More by Temur Rustamovich Khudayberganov
- Industry Practitioners Developing or Implementing AI Platforms for Healthcare and Life SciencesAdvances in computational intelligence and robotics book series · 2026
- Hybrid DES and Continuous Models for Cyber-Physical Ubiquitous EnvironmentsAdvances in computational intelligence and robotics book series · 2026
- Methods for Geometrically Robust Encoding and Synchronization of Audio-Video Metadata in Parametric 3D Surfaces (NURBS/Mesh): Taxonomy, Threat Models, Design Requirements, and Benchmark EvidenceInternational Siberian workshops and tutorials on electron devices and materials · 2026
- Security-Aware Geometrically Robust 3D Watermarking: Threat Models, Games, and Evaluation ProtocolsInternational Siberian workshops and tutorials on electron devices and materials · 2026
