LiSTA: Geometric Object-Based Change Detection in Cluttered Environments

TL;DR

LiSTA (LiDAR Spatio-Temporal Analysis), a system to detect probabilistic object-level change over time using multi-mission LiDAR SLAM, demonstrates superior performance in detecting changes in semi-static environments compared to existing methods.

Abstract

We present LiSTA (LiDAR Spatio-Temporal Analysis), a system to detect probabilistic object-level change over time using multi-mission SLAM. Many applications require such a system, including construction, robotic navigation, long-term autonomy, and environmental monitoring. We focus on the semi-static scenario where objects are added, subtracted, or changed in position over weeks or months. Our system combines multi-mission LiDAR SLAM, volumetric differencing, object instance description, and correspondence grouping using learned descriptors to keep track of an open set of objects. Object correspondences between missions are determined by clustering the object's learned descriptors. We demonstrate our approach using datasets collected in a simulated environment and a real-world dataset captured using a LiDAR system mounted on a quadruped robot monitoring an industrial facility containing static, semi-static, and dynamic objects. Our method demonstrates superior performance in detecting changes in semi-static environments compared to existing methods.

Figures (8)

• Figure 1: Top down view of 3D maps acquired by an autonomous Spot inspecting Fire Service College, static map in blue; with changes identified between missions A and B in red. Photos show our in-house sensor suite Frontier and Spot quadruped.
• Figure 2: LiSTA change detection method overview. Five modules process the sensor data to output the classified changed objects intermission.
• Figure 3: Multi-Mission SLAM registration factor graph structure: It identifies loop closures between missions ($M_1$, $M_2$ & $M_3$) and jointly optimizes their pose-graphs.
• Figure 4: Examples of segmented object point cloud clusters of the same class, after unsupervised classification. Top: chairs from the simulated LiDAR dataset. Bottom: cars from the real-world FSC experiment. This shows the robustness of the classifier to minor occlusion and different viewpoints.
• Figure 5: Comparison between (a) naive direct ICP alignment and (b) our multi-mission SLAM method. (a) Shows the cloud-to-cloud distance of a local part of the mild misregisteration between two rigid global point clouds, "double walling" occurs locally causing phantom change to be detected. (b) Shows the cloud to cloud distance of our multi-mission registration solving this problem.