Autonomous vision-based UAV system for oil spill sampling and precision recovery

Document Type

Article

Publication Date

6-11-2026

Department

Department of Applied Computing

Abstract

Oil spills pose a major environmental and economic threat to marine and coastal ecosystems. Rapid detection and containment are critical, as delays can allow spills to spread, increasing ecological damage and cleanup costs. Different oils require different response strategies, including mechanical recovery for heavy oils, natural dispersion for light oils, and heating or specialized pumps for viscous oils. Without sampling, responders cannot choose the correct mitigation approach. This work presents the development of a vision-based autonomous UAV module that enables oil spill detection, sample collection, and autonomous recovery, addressing the need for rapid unmanned incident response. The proposed module is platform-agnostic and integrates onboard perception, navigation, and control to support visual target recognition, autonomous mission execution, and precision landing. The system autonomously navigates, performs simulated sampling, and executes precision landing guided by fiducial landing Apriltags. The AprilTag2 package within the Robot Operating System (ROS) framework is used to detect fiducial tags in images captured by an Intel RealSense D435 camera mounted on the drone. A statemachine- based landing algorithm is developed to enable precise landing on detected AprilTags. The proposed algorithm computes the tag’s center as the target landing point, transforms its coordinates from the camera frame to the robot frame, and guides the drone to land precisely on the detected tag using a PID controller. A proof-of-concept experiment demonstrated the full operational cycle: the UAV autonomously takes off, navigates to a GPS waypoint representing the spill site (which is represented by an Apriltag) simulating the spill, the drone descends to a predefined altitude for mock sampling, and returns to the home platform for precision landing (which is represented by a second Apriltag). Additionally, a lightweight passive sampling mechanism is proposed for UAV-based oil spill response, enabling in-situ collection without additional power or actuation. The design uses a self-actuating flap system to capture and retain samples while minimizing impact on payload and flight time.

Publication Title

Proc. SPIE 14032, Pattern Recognition and Prediction XXXVII

Link to Full Text

Share

COinS