Real-Time Human Detection and Gesture Recognition for On-Board UAV Rescue

Area of Science:

• Robotics and Artificial Intelligence
• Search and Rescue Technology
• Human-Computer Interaction

Background:

• Unmanned aerial vehicles (UAVs) are increasingly vital in technical fields, particularly for wilderness rescue operations.
• Effective communication between rescue personnel and individuals in distress is crucial but often challenging in remote environments.
• Existing communication methods can be limited by environmental factors or the condition of the rescued individual.

Purpose of the Study:

• To develop a real-time human detection and rescue gesture recognition system for UAVs.
• To enable biometric communication between humans and UAVs using body and hand gestures.
• To enhance the efficiency and effectiveness of UAV-assisted wilderness rescue operations.

Main Methods:

• Utilized YOLOv3-tiny for efficient human detection in real-time.
• Developed a dataset of ten distinct body rescue gestures captured by a UAV's onboard camera.
• Implemented deep learning models for recognizing both body and hand gestures, including novel dynamic gestures like 'Attention' and 'Cancel'.

Main Results:

• Achieved 99.80% accuracy in recognizing body rescue gestures from the created dataset.
• Attained 94.71% accuracy in recognizing hand gestures.
• Demonstrated successful real-time performance using UAV onboard cameras, confirming the system's rescue capabilities.

Conclusions:

• The developed UAV system effectively detects humans and recognizes rescue gestures, facilitating non-verbal communication.
• The system's high accuracy in gesture recognition supports its application in critical wilderness rescue scenarios.
• This technology offers a promising solution for improving UAV-based search and rescue missions.