面向三维异常检测的几何对齐与异常感知重建
Geometry-Aligned and Anomaly-Aware Reconstruction for 3D Anomaly Detection
把扩散去噪的噪声方向约束到顶点法向上、并用自适应 Transformer 按区域分配参考，是可直接迁移到点云补全、去噪与表面重建的几何先验注入方式；对做 3D 重建的人来说，它给出了「噪声采样服从局部几何」这一可复用设计。
原文依据 AARD progressively aligns noise directions with vertex normals while maintaining vertex-graph consistency
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论点云异常检测在自动化制造中至关重要，基于重建的扩散方法已成为主流方案。然而，这类方法仍面临两大挑战：(1) 几何违背，即随机噪声扰动偏离局部表面法向，导致结构畸变；(2) 参考区域未加区分，即在去噪过程中统一施加粗糙的异常嵌入，会模糊正常细节并妨碍准确的异常恢复。为解决这些问题，我们提出 AARD，一个几何对齐且异常感知的扩散重建框架。我们认为，高保真异常检测需要对扩散过程进行有原则的重构：噪声应与几何对齐以保持结构，重建可由异常感知的参考更好地引导，从而有区分地恢复正常细节并修复缺陷。AARD 在保持顶点图一致性的同时，逐步将噪声方向与顶点法向对齐，并采用自适应 Transformer，为异常区域分配正常参考、为正常区域分配输入参考。在 Anomaly-ShapeNet 和 Real3D-AD 上的实验表明，AARD 持续优于最优方法，实现了更优的几何保真度与稳健的异常定位。
英文原文摘要 1334 字符
Point cloud anomaly detection is crucial in automated manufacturing, with reconstruction-based diffusion methods emerging as a mainstream solution. However, these approaches still face two major challenges: (1) geometry violation, where random noise perturbations deviate from local surface normals, causing structural distortion; and (2) undistinguished reference regions, where uniformly applied coarse anomaly embeddings during denoising blur normal details and impede accurate anomaly recovery. To address these issues, we propose AARD, a geometry-aligned and anomaly-aware diffusion reconstruction framework. We argue that high-fidelity anomaly detection requires a principled reformulation of the diffusion process: noise should align with geometry to preserve structures, and reconstruction can be better guided by anomaly-aware references to discriminatively recover normal details while correcting defects. AARD progressively aligns noise directions with vertex normals while maintaining vertex-graph consistency, and employs an adaptive transformer that assigns normal references to anomalous regions and input references to normal areas. Experiments on Anomaly-ShapeNet and Real3D-AD show that AARD consistently outperforms state-of-the-art approaches, achieving superior geometric fidelity and robust anomaly localization.