Obstacle detection for lake-deployed autonomous surface vehicles using RGB imagery.

Area of Science:

• Robotics
• Computer Vision
• Environmental Monitoring

Background:

• Autonomous surface vehicles (ASVs) require robust obstacle detection for safe lake operations.
• Variable lake environments present challenges such as shorelines, land-sky boundaries, and glint.
• Existing systems may be too costly or complex for small, lake-deployed ASVs.

Purpose of the Study:

• To develop and test an effective, low-cost obstacle detection system for small ASVs operating in lakes.
• To address the unique challenges of lake environments, including shoreline presence and variable lighting.
• To demonstrate the feasibility of using single-board computers for real-time ASV obstacle detection.

Main Methods:

• An innovative gradient-based image processing algorithm was developed for low-angle camera views.
• Correlation-based multi-frame analysis was employed to differentiate true and false positive obstacle detections.
• The system was implemented on a Raspberry Pi 3 Model B computer using a consumer-grade camera.

Main Results:

• The obstacle detection algorithm processed images at approximately 3-4 Hz under operational conditions.
• Extensive testing on Lake Geneva confirmed the system's effectiveness in variable lake conditions.
• The system successfully demonstrated obstacle detection despite shoreline interference and glint.

Conclusions:

• Single-board computers are suitable for developing effective and low-cost obstacle detection systems for ASVs.
• The developed algorithm provides a viable solution for autonomous navigation in complex lake environments.
• This technology enhances the safety and operational capabilities of small ASVs.