A 3D Multimodal Feature for Infrastructure Anomaly Detection

2025-02-09

AI Summary

Detecting subtle cracks and moisture-induced anomalies in aging masonry bridge point clouds remains challenging due to their low geometric contrast and sparse intensity variations. To address this, we propose 3DMulti-FPFHI, an unsupervised 3D multimodal anomaly detection method. It introduces intensity-enhanced local features—uniquely unifying LiDAR intensity anomalies with geometric anomalies—and integrates customized Fast Point Feature Histograms (FPFH) with raw intensity values. Built upon the PatchCore framework, it enables efficient, few-shot anomaly detection using only a small set of unlabeled, normal-point-cloud samples. Evaluated on real-world stone-arch bridge and concrete tunnel point clouds, 3DMulti-FPFHI significantly improves detection recall for sub-centimeter cracks and accurately identifies intensity degradation caused by water seepage. Quantitative results demonstrate superior precision and robustness over conventional FPFH-based methods and state-of-the-art multimodal anomaly detectors.

Technology Category

• Computer Vision: Multi-modal Vision
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Multimodal Learning

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Ageing structures require periodic inspections to identify structural defects. Previous work has used geometric distortions to locate cracks in synthetic masonry bridge point clouds but has struggled to detect small cracks. To address this limitation, this study proposes a novel 3D multimodal feature, 3DMulti-FPFHI, that combines a customized Fast Point Feature Histogram (FPFH) with an intensity feature. This feature is integrated into the PatchCore anomaly detection algorithm and evaluated through statistical and parametric analyses. The method is further evaluated using point clouds of a real masonry arch bridge and a full-scale experimental model of a concrete tunnel. Results show that the 3D intensity feature enhances inspection quality by improving crack detection; it also enables the identification of water ingress which introduces intensity anomalies. The 3DMulti-FPFHI outperforms FPFH and a state-of-the-art multimodal anomaly detection method. The potential of the method to address diverse infrastructure anomaly detection scenarios is highlighted by the minimal requirements for data compared to learning-based methods. The code and related point cloud dataset are available at https://github.com/Jingyixiong/3D-Multi-FPFHI.

Problem

• Detect small cracks in ageing structures
• Improve infrastructure anomaly detection quality
• Combine FPFH with intensity for better results

Innovation

• Combines FPFH with intensity feature
• Integrates into PatchCore algorithm
• Enhances crack and water ingress detection