Year: 2025
Type: conference-paper
Abstract: This study developed an obstacle avoidance system for drones using depth estimation from RGB cameras, aiming to reduce reliance on expensive sensors like RGB-D cameras or LiDAR.The system employs the deep learning model ZoeDepth for depth estimation and integrates it with ROS and Gazebo for simulation.Two autonomous systems were evaluated: one using RGB-D cameras and the other using depth estimation with RGB cameras.Experimental results show that while the RGB-D camera system outperformed in accuracy, the depth estimation-based system provided cost-effective and reasonable performance, especially in complex environments.The research concludes with plans to improve the system for denser obstacle environments and conduct real-world experiments.
Authors: Sora Takahashi, Eiji Hayashi
Institution: Kyushu Institute of Technology
Language: English
Topic: Advanced Vision and Imaging
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4409914129