Evaluating the Forest Ecosystem through a Semi-Autonomous Quadruped Robot and a Hexacopter UAV

Area of Science:

• Forestry science
• Robotics
• Computer vision

Background:

• Forest resilience is crucial for economic growth and climate regulation, necessitating accurate monitoring.
• Citizen science in the UK aids forest management but faces challenges with tedious data collection.
• Ground-level forest health indicators (FHIs) are often inaccessible via traditional aerial surveys.

Purpose of the Study:

• To develop an efficient system for detecting ground-level FHIs inaccessible to aerial surveys.
• To evaluate object detection algorithms for real-time, low-latency FHIs detection.
• To enable cost-effective and autonomous forest monitoring.

Main Methods:

• Deployed a quadruped robot with an integrated camera and an external sensing platform (ESP) with light and infrared cameras.
• Analyzed various YOLO object-detection algorithm versions for accuracy, deployment, and usability.
• Created new datasets to train YOLOv5x and YOLOv5s for FHIs recognition.

Main Results:

• YOLOv5s demonstrated lightweight characteristics and ease of training for FHI detection.
• YOLOv5s achieved near real-time performance, suitable for autonomous monitoring.
• The system enables cost-effective and autonomous forest health monitoring.

Conclusions:

• The quadruped robot and ESP system effectively detect ground-level FHIs.
• YOLOv5s is a viable algorithm for efficient and autonomous forest health monitoring.
• This approach enhances forest management capabilities through advanced technology.