Point cloud upsampling improves 3D anomaly detection on low-resolution sensors
PC$^2$-AD: Point Cloud Upsampling to Safeguard 3D Anomaly Detection with Resolution-constrained Edge Devices
Computer Vision and Pattern Recognition
Summary
Using cheap sensors at the edge can create sparse 3D data that makes it harder to find problems with objects. The authors introduce a method called PC²-AD that adds points to sparse 3D scans before checking for anomalies. Their system chooses extra points carefully to match the original shapes and keep normal parts intact. Tests show this approach helps many existing anomaly detectors work better with low-quality 3D inputs.
What this means in practice
- •For edge system developers: Enhance 3D anomaly detection accuracy when using affordable, low-resolution sensors by compensating sparse point clouds before analysis.
- •For industrial inspection teams: Improve defect detection reliability on 3D scans collected from low-cost devices by adding quality-aware upsampling as preprocessing.
Authors
Yutong Gu, Yingxi Xie, Kejin Huang, Jian Ning, Hanzhe Liang, Linlin Shen, Jinbao Wang
Abstract
Low-cost and low-resolution sensors used in edge deployments can produce test point clouds that are substantially sparser than the normal training data. This train-test sampling-resolution gap changes the local geometry available to a 3D anomaly detector. We propose PC$^2$-AD, a point cloud upsampling framework that compensates sparse test inputs before downstream detection. Target Domain Candidate Generation (TCG) adapts a pretrained upsampler to normal training geometry and generates a dense candidate pool. Geometry-Aware Candidate Filtering (GACF) selects candidates according to geometric spacing and spatial coverage. Normality-Preserving Point Compensation (NPPC) refines the selection by comparing candidate normality scores with those of their input anchors. The selected points are combined with the unchanged input points and processed by the existing detector. Experiments with six detectors on two Anomaly-ShapeNet settings and Real3D-AD show improvements in the mean of object-level and point-level AUROC for all six detectors in each Anomaly-ShapeNet setting and four on Real3D-AD. These results support point cloud compensation as an input-level approach to improving 3D anomaly detection under low-resolution sensing conditions. Code is publicly available at https://github.com/gyutong406-commits/PC2-AD.