3D-PNAS: 3D Industrial Surface Anomaly Synthesis with Perlin Noise

TL;DR

This work addresses the limited availability of labeled 3D industrial defects by introducing 3D-PNAS, a Perlin-noise–based method that deforms 3D point clouds along surface normals after projecting onto a 2D plane via PCA. It sampling a continuous 2D noise field with tunable scale, octaves, persistence, and lacunarity, then applies adaptive, thresholded, and locally normalized displacements to generate physically plausible anomalies with controllable size and detail. Key contributions include a simple yet effective 3D anomaly generator, a complete codebase and visualization toolkit, and cross-category demonstrations showing consistency across diverse object geometries. The approach enables targeted data augmentation for 3D anomaly detection and provides a foundation for integrating synthetic 3D defects with pretrained foundation models in industrial inspection contexts.

Abstract

Large pretrained vision foundation models have shown significant potential in various vision tasks. However, for industrial anomaly detection, the scarcity of real defect samples poses a critical challenge in leveraging these models. While 2D anomaly generation has significantly advanced with established generative models, the adoption of 3D sensors in industrial manufacturing has made leveraging 3D data for surface quality inspection an emerging trend. In contrast to 2D techniques, 3D anomaly generation remains largely unexplored, limiting the potential of 3D data in industrial quality inspection. To address this gap, we propose a novel yet simple 3D anomaly generation method, 3D-PNAS, based on Perlin noise and surface parameterization. Our method generates realistic 3D surface anomalies by projecting the point cloud onto a 2D plane, sampling multi-scale noise values from a Perlin noise field, and perturbing the point cloud along its normal direction. Through comprehensive visualization experiments, we demonstrate how key parameters - including noise scale, perturbation strength, and octaves, provide fine-grained control over the generated anomalies, enabling the creation of diverse defect patterns from pronounced deformations to subtle surface variations. Additionally, our cross-category experiments show that the method produces consistent yet geometrically plausible anomalies across different object types, adapting to their specific surface characteristics. We also provide a comprehensive codebase and visualization toolkit to facilitate future research.

Figures (7)

• Figure 1: Comparison between cut-paste based approach and our approach in 3D anomaly synthesis.
• Figure 2: Overview of our 3D anomaly synthesis method: (a) Original point cloud with PCA directions; (b) Projected points on reference plane; (c) Perlin noise space; and (d) Augmented point cloud with anomalies. This pipeline ensures that the generated anomalies respect the local surface geometry while providing diverse and controllable deformations.
• Figure 3: Visualization of the effect of varying the noise scale ($s$) on the generated anomalies in a 3D point cloud: (a) $s=1.0$ produces large, smooth, and sparse deformations; (b) $s=2.0$ creates medium-sized deformations; and (c) $s=4.0$ results in smaller, more localized, and densely distributed anomalies. Red points indicate protrusions, while blue points indicate intrusions.
• Figure 4: Visualization of the effect of varying the perturbation strength ($\alpha$) on the generated anomalies in a 3D point cloud: (a) $\alpha=0.01$ creates subtle surface modifications; (b) $\alpha=0.02$ produces moderate deformations; and (c) $\alpha=0.05$ results in more pronounced and significant protrusions and intrusions. All cases maintain the same spatial distribution pattern while varying in magnitude.
• Figure 5: Visualization of the effect of varying the octaves ($o$) on the generated anomalies in a 3D point cloud: (a) $o=1$ produces smoother, single-scale anomalies; (b) $o=2$ adds an intermediate level of detail; and (c) $o=4$ results in more complex, multi-scale anomalies with both large-scale structures and fine-grained details.