Multi-Sensor Object Anomaly Detection: Unifying Appearance, Geometry, and Internal Properties

TL;DR

This paper introduces MulSen-AD, a multi-sensor dataset and benchmark for industrial object anomaly detection, and proposes a fusion algorithm that significantly improves detection accuracy over single-sensor methods.

Contribution

The paper presents the first multi-sensor dataset, benchmark, and fusion algorithm for comprehensive industrial object anomaly detection.

Findings

Multi-sensor fusion achieves 96.1% AUROC in detection accuracy.

The dataset covers 15 industrial products with diverse anomalies.

Fusion outperforms single-sensor approaches significantly.

Abstract

Object anomaly detection is essential for industrial quality inspection, yet traditional single-sensor methods face critical limitations. They fail to capture the wide range of anomaly types, as single sensors are often constrained to either external appearance, geometric structure, or internal properties. To overcome these challenges, we introduce MulSen-AD, the first high-resolution, multi-sensor anomaly detection dataset tailored for industrial applications. MulSen-AD unifies data from RGB cameras, laser scanners, and lock-in infrared thermography, effectively capturing external appearance, geometric deformations, and internal defects. The dataset spans 15 industrial products with diverse, real-world anomalies. We also present MulSen-AD Bench, a benchmark designed to evaluate multi-sensor methods, and propose MulSen-TripleAD, a decision-level fusion algorithm that integrates these three modalities for robust, unsupervised object anomaly detection. Our experiments demonstrate that multi-sensor fusion substantially outperforms single-sensor approaches, achieving 96.1% AUROC in object-level detection accuracy. These results highlight the importance of integrating multi-sensor data for comprehensive industrial anomaly detection.