Vision-Based Target Finding and Inspection of a Ground Target Using a Multirotor UAV System

Area of Science:

• Robotics
• Aerospace Engineering
• Computer Vision

Background:

• Automated aerial systems are increasingly vital for complex tasks.
• Precise navigation and target identification are critical for Unmanned Aerial Vehicle (UAV) applications.
• Existing systems may lack robustness against environmental and sensor-related challenges.

Purpose of the Study:

• To develop and evaluate an integrated system for autonomous ground target detection, inspection, and navigation using a multirotor UAV.
• To enable applications such as precise payload delivery and site-specific crop management.
• To create a robust system capable of handling sensor errors and external disturbances.

Main Methods:

• A multirotor Unmanned Aerial Vehicle (UAV) equipped with a downward-looking camera for target detection.
• Implementation of a high-level decision algorithm based on the Observe, Orient, Decide, Act (OODA) loop.
• Mavros for real-time navigation by sending local position messages to the autopilot.
• System tested in simulations and outdoor trials from 10m to 40m.

Main Results:

• The system successfully performed target detection, descent, and hovering above the target.
• Demonstrated high reliability and robustness against sensor errors, drift, and external disturbances.
• Validated performance across multiple trials in both simulated and real-world environments.

Conclusions:

• The developed UAV system offers a reliable and robust solution for autonomous target detection and inspection.
• The OODA loop-based algorithm effectively guides the UAV through locate, descend, and hover phases.
• The system shows significant potential for applications in precision agriculture and logistics.