Reimagining service delivery in the agentic era with Google Public Sector
cloud.google.com·19h ago
AI Research
arXiv CS.AI·Jun 11·5 min read
EKF-Based Depth Camera and Deep Learning Fusion for UAV-Person Distance Estimation and Following in SAR Operations
TL;DR
Unmanned Aerial Vehicles (UAVs) face challenges in accurately estimating the distance to individuals during search and rescue operations. This study developed an Extended Kalman Filter (EKF) based method that fuses depth camera data with deep learning techniques for improved distance estimation.
✦ Why It Matters
Engineers can leverage EKF and deep learning fusion techniques to enhance UAV safety and effectiveness in real-world tracking scenarios.
Key Takeaways
- Utilize EKF for real-time data fusion in UAV applications.
- Incorporate YOLO-pose for human body keypoint detection.
- Test systems in varied environments to ensure robustness.
- Aim for reduced distance estimation errors in tracking.
How It Works
The system employs a depth camera to gather spatial data and a monocular camera to estimate the distance to human targets. YOLO-pose processes the monocular images to identify keypoints on the human body, while the EKF algorithm fuses this information with depth data to provide a continuous and accurate distance measurement.
This fusion allows the UAV to maintain a safe following distance, adapting to real-time changes in the environment.