3D Point Cloud Data

We propose 3D-CSAD, a novel method for untrained 3D anomaly detection in complex manufacturing surfaces using 3D point cloud data. Our method segments the surface into simpler components and uses Robust Principal Component Analysis (RPCA) for accurate detection.

This paper presents PointSGRADE, a novel sparse learning framework with graph representation for anomaly detection on smooth free-form surfaces using unstructured 3D point cloud data. The methodology addresses challenges such as irregular data structures and variant anomaly patterns, offering a computationally efficient solution.

This paper proposes a novel Bayesian network-based methodology for anomaly detection in fabricated artifacts using unstructured 3D point cloud data. The approach addresses challenges such as nonexistence of global coordinate ordering, variant anomaly patterns, and the presence of outliers.

This paper proposes a tensor voting-based approach for classifying surface anomalies on artifacts using 3D point cloud data, addressing challenges such as complex data representation, high dimensionality, and inconsistent point sizes.