This repository contains ROS packages for ESP-drone camera image processing and ArUco marker detection and pose tracking.
This stack works in conjunction with espdrone-ros, a set of ROS packages for interfacing with the ESP-drone using ROS, including for fetching camera image from the drone. Both this repository and espdrone-ros are specialized for the NTU EEE UAVIONICS DIP project.
This stack as well as aruco-3.1.12, the required dependency not in this repository, have been tested on Ubuntu 20.04 and ROS Noetic (both using Docker and native system installation).
The packages in this repository allow for processing the camera image from multiple ESP-drones concurrently, specifically to:
- Process the camera image using ROS
image_proc, e.g. for camera undistortion and convertion to grayscale (MONO8) format - Perform pose estimation (position and orientation) of the camera/drone using multiple ArUco fiducial markers scattered across the drone's environment, forming a marker map
Pose estimation is possible if at least 2 ArUco markers are in the field of view of the drone's camera. This is the current limitation of the aruco-3.1.12 library.
This stack is CPU bound; no powerful GPU is required. The stack has been tested to run fine on a Raspberry Pi 4B (8GB model) running Ubuntu 20.04 MATE, though a computer with a more powerful CPU is highly recommended if you intend to launch more than one ESP-drones.
The computer running this stack also needs to connect to the same network as the ESP-drone.
For guides on setup and usage or notes on this stack, please head to the Wiki section of this repository.