Registration is a Powerful Rotation-Invariance Learner for 3D Anomaly Detection

2025-10-19

AI Summary

To address weak feature discriminability, insufficient local geometric modeling, and poor rotation invariance in 3D point cloud anomaly detection, this paper proposes a registration-driven rotation-invariant feature extraction framework. The method jointly optimizes geometric alignment and representation learning by leveraging the point cloud registration process as a supervisory signal—a novel use of registration for feature learning. It further integrates a registration network with a memory bank mechanism to enhance local structural modeling and cross-sample discriminability. Evaluated on Anomaly-ShapeNet and Real3D-AD, the approach significantly outperforms existing state-of-the-art methods. Notably, it maintains high robustness and generalization under challenging conditions—including arbitrary rotations and registration failures—demonstrating superior geometric consistency and adaptability.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Computer Vision: 3D Computer Vision
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

3D anomaly detection in point-cloud data is critical for industrial quality control, aiming to identify structural defects with high reliability. However, current memory bank-based methods often suffer from inconsistent feature transformations and limited discriminative capacity, particularly in capturing local geometric details and achieving rotation invariance. These limitations become more pronounced when registration fails, leading to unreliable detection results. We argue that point-cloud registration plays an essential role not only in aligning geometric structures but also in guiding feature extraction toward rotation-invariant and locally discriminative representations. To this end, we propose a registration-induced, rotation-invariant feature extraction framework that integrates the objectives of point-cloud registration and memory-based anomaly detection. Our key insight is that both tasks rely on modeling local geometric structures and leveraging feature similarity across samples. By embedding feature extraction into the registration learning process, our framework jointly optimizes alignment and representation learning. This integration enables the network to acquire features that are both robust to rotations and highly effective for anomaly detection. Extensive experiments on the Anomaly-ShapeNet and Real3D-AD datasets demonstrate that our method consistently outperforms existing approaches in effectiveness and generalizability.

Problem

• Achieving rotation-invariant 3D anomaly detection in point clouds
• Addressing inconsistent feature transformations in memory bank methods
• Integrating registration with feature learning for reliable defect identification

Innovation

• Registration-induced rotation-invariant feature extraction framework
• Joint optimization of alignment and representation learning
• Integration of registration and memory-based anomaly detection