Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Clustering-based Learning for UAV Tracking and Pose Estimation

Jiaping Xiao, Phumrapee Pisutsin, Cheng Wen Tsao, Mir Feroskhan

TL;DR

This work addresses robust 3D UAV tracking and pose estimation under multimodal sensing by fusing LiDAR360 and Livox Avia data. It proposes CL-Det, a clustering-based approach that aligns sensor timestamps, isolates objects-of-interest, and localizes UAVs in 3D using DBSCAN, with a fallback to historical estimations when data are missing. The method achieves competitive performance on the CVPR 2024 UG2+ Track 5 dataset, ranking 5th and delivering around 14.9 predictions per second using LiDAR data. The contribution demonstrates how clustering and multi-sensor fusion can enable reliable 3D UAV tracking in partially observed environments, with practical implications for formation control and anti-UAV systems.

Abstract

UAV tracking and pose estimation plays an imperative role in various UAV-related missions, such as formation control and anti-UAV measures. Accurately detecting and tracking UAVs in a 3D space remains a particularly challenging problem, as it requires extracting sparse features of micro UAVs from different flight environments and continuously matching correspondences, especially during agile flight. Generally, cameras and LiDARs are the two main types of sensors used to capture UAV trajectories in flight. However, both sensors have limitations in UAV classification and pose estimation. This technical report briefly introduces the method proposed by our team "NTU-ICG" for the CVPR 2024 UG2+ Challenge Track 5. This work develops a clustering-based learning detection approach, CL-Det, for UAV tracking and pose estimation using two types of LiDARs, namely Livox Avia and LiDAR 360. We combine the information from the two data sources to locate drones in 3D. We first align the timestamps of Livox Avia data and LiDAR 360 data and then separate the point cloud of objects of interest (OOIs) from the environment. The point cloud of OOIs is clustered using the DBSCAN method, with the midpoint of the largest cluster assumed to be the UAV position. Furthermore, we utilize historical estimations to fill in missing data. The proposed method shows competitive pose estimation performance and ranks 5th on the final leaderboard of the CVPR 2024 UG2+ Challenge.

Clustering-based Learning for UAV Tracking and Pose Estimation

TL;DR

This work addresses robust 3D UAV tracking and pose estimation under multimodal sensing by fusing LiDAR360 and Livox Avia data. It proposes CL-Det, a clustering-based approach that aligns sensor timestamps, isolates objects-of-interest, and localizes UAVs in 3D using DBSCAN, with a fallback to historical estimations when data are missing. The method achieves competitive performance on the CVPR 2024 UG2+ Track 5 dataset, ranking 5th and delivering around 14.9 predictions per second using LiDAR data. The contribution demonstrates how clustering and multi-sensor fusion can enable reliable 3D UAV tracking in partially observed environments, with practical implications for formation control and anti-UAV systems.

Abstract

UAV tracking and pose estimation plays an imperative role in various UAV-related missions, such as formation control and anti-UAV measures. Accurately detecting and tracking UAVs in a 3D space remains a particularly challenging problem, as it requires extracting sparse features of micro UAVs from different flight environments and continuously matching correspondences, especially during agile flight. Generally, cameras and LiDARs are the two main types of sensors used to capture UAV trajectories in flight. However, both sensors have limitations in UAV classification and pose estimation. This technical report briefly introduces the method proposed by our team "NTU-ICG" for the CVPR 2024 UG2+ Challenge Track 5. This work develops a clustering-based learning detection approach, CL-Det, for UAV tracking and pose estimation using two types of LiDARs, namely Livox Avia and LiDAR 360. We combine the information from the two data sources to locate drones in 3D. We first align the timestamps of Livox Avia data and LiDAR 360 data and then separate the point cloud of objects of interest (OOIs) from the environment. The point cloud of OOIs is clustered using the DBSCAN method, with the midpoint of the largest cluster assumed to be the UAV position. Furthermore, we utilize historical estimations to fill in missing data. The proposed method shows competitive pose estimation performance and ranks 5th on the final leaderboard of the CVPR 2024 UG2+ Challenge.
Paper Structure (16 sections, 7 figures, 1 table)

This paper contains 16 sections, 7 figures, 1 table.

Table of Contents

  1. Introduction
  2. Exploratory Work
  3. K-Means Clustering Approach
  4. Optimization of Cluster Parameters
  5. Monitoring and Estimating Drone Position
  6. Parameter Tuning
  7. Elbow Method Application
  8. Methodology
  9. Data Sources
  10. Algorithm
  11. LiDAR 360 Data Processing
  12. Livox Avia Data Processing
  13. Fallback Method
  14. Implementation Details
  15. Results
  16. ...and 1 more sections

Figures (7)

  • Figure 1: Sample LiDAR 360 3D point cloud data plot
  • Figure 2: Sample Livox Avia 3D point cloud data plot
  • Figure 3: Heatmap of the drone position ground truth
  • Figure 4: Histogram of the drone position ground truth
  • Figure 5: 2D plots of LiDAR 360 point cloud with environment data point
  • ...and 2 more figures