Finding Outliers in a Haystack: Anomaly Detection for Large Pointcloud Scenes

TL;DR

This paper introduces a novel reconstruction-based open-set segmentation approach for large-scale outdoor LiDAR point clouds, improving detection of outliers and integrating with existing methods using a scalable Mamba architecture.

Contribution

The paper presents a new reconstruction-based method for outdoor scene open-set segmentation and a Mamba architecture that enhances scalability and performance on large point clouds.

Findings

Improved open-set segmentation performance on large point clouds.

Mamba architecture is competitive with voxel-convolution methods.

Method enhances outlier detection in outdoor LiDAR data.

Abstract

LiDAR scanning in outdoor scenes acquires accurate distance measurements over wide areas, producing large-scale point clouds. Application examples for this data include robotics, automotive vehicles, and land surveillance. During such applications, outlier objects from outside the training data will inevitably appear. Our research contributes a novel approach to open-set segmentation, leveraging the learnings of object defect-detection research. We also draw on the Mamba architecture's strong performance in utilising long-range dependencies and scalability to large data. Combining both, we create a reconstruction based approach for the task of outdoor scene open-set segmentation. We show that our approach improves performance not only when applied to our our own open-set segmentation method, but also when applied to existing methods. Furthermore we contribute a Mamba based architecture which is competitive with existing voxel-convolution based methods on challenging, large-scale pointclouds.