A Lightweight 3D Anomaly Detection Method with Rotationally Invariant Features

2025-11-17

AI Summary

To address the degradation of 3D point cloud anomaly detection performance under pose variations due to feature instability, this paper proposes RIF, a rotation-invariant lightweight anomaly detection framework. Our approach tackles the problem through three key innovations: (1) Point Coordinate Mapping (PCM), which transforms raw 3D coordinates into rotation-invariant representations; (2) CTF-Net, a lightweight convolutional transformation network leveraging transfer learning and 3D-specific data augmentation to enhance feature discriminability; and (3) a memory bank mechanism enabling efficient and scalable anomaly scoring. Evaluated on Anomaly-ShapeNet and Real3D-AD benchmarks, RIF achieves absolute improvements of +17.7% and +1.6% in pixel-level AUROC (P-AUROC), respectively, demonstrating significantly enhanced generalization across diverse object poses. The framework’s computational efficiency and robustness make it suitable for practical industrial deployment.

Technology Category

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

Application Category

• Responsible Web: Algorithmic accountability and transparency on the web
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

3D anomaly detection (AD) is a crucial task in computer vision, aiming to identify anomalous points or regions from point cloud data. However, existing methods may encounter challenges when handling point clouds with changes in orientation and position because the resulting features may vary significantly. To address this problem, we propose a novel Rotationally Invariant Features (RIF) framework for 3D AD. Firstly, to remove the adverse effect of variations on point cloud data, we develop a Point Coordinate Mapping (PCM) technique, which maps each point into a rotationally invariant space to maintain consistency of representation. Then, to learn robust and discriminative features, we design a lightweight Convolutional Transform Feature Network (CTF-Net) to extract rotationally invariant features for the memory bank. To improve the ability of the feature extractor, we introduce the idea of transfer learning to pre-train the feature extractor with 3D data augmentation. Experimental results show that the proposed method achieves the advanced performance on the Anomaly-ShapeNet dataset, with an average P-AUROC improvement of 17.7%, and also gains the best performance on the Real3D-AD dataset, with an average P-AUROC improvement of 1.6%. The strong generalization ability of RIF has been verified by combining it with traditional feature extraction methods on anomaly detection tasks, demonstrating great potential for industrial applications.

Problem

• Detecting anomalies in 3D point cloud data with varying orientations
• Maintaining feature consistency despite positional and rotational changes
• Developing lightweight robust features for industrial anomaly detection

Innovation

• Rotationally invariant features framework for 3D anomaly detection
• Point coordinate mapping technique for consistent representation
• Lightweight convolutional transform feature network for robust features