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

Lei Zhang (International Digital Economy Academy (IDEA)) · Shengfeng He (Singapore Management University) · Yuyang Yu (South China University of Technology) · Zhengwei Chen (South China University of Technology) · Xuemiao Xu (South China University of Technology) · Haoxin Yang (South China University of Technology) · Yongwei Nie (South China University of Technology)

3d anomaly detectionalignment optimizationanomaly-shapenetdiscriminative capacityfeature extraction frameworkfeature similarityfeature transformationslocal geometric detailsmemory bank-based methodspoint-cloud datapoint-cloud registrationreal3d-ad datasetsrepresentation learningrotation invariancerotation-invariant representations

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.