EKF-Based Depth Camera and Deep Learning Fusion for UAV-Person Distance Estimation and Following in SAR Operations
Researchers have developed a new system for Unmanned Aerial Vehicles (UAVs) that fuses depth camera data with deep learning techniques to accurately estimate and maintain a safe distance from individuals. This approach utilizes the Extended Kalman Filter (EKF) algorithm to combine depth information with monocular camera-based distance estimations, enhanced by YOLO-pose for real-time processing. The system has demonstrated improved accuracy and robustness in challenging conditions, reducing distance estimation errors by up to 15.3% in tested scenarios and extending the effective depth detection range. AI
IMPACT Enhances safety and precision for AI-driven autonomous systems in critical search and rescue operations.