This repo implements the visual positioning & controlling functionalities for the drone delivery platform.
In brief, the program detects the aruco tag captured by the PiCam, calculates the relative position between it and the drone, and use mavlink protocol to navigate PixHawk-based drone to land on the tag safely.
- Generation of Aruco tags(this is emitted, just use some tools)
- Detection of Aruco tags
- Positioning of Aruco tags in 3D space
- Pose estimation
- Distance estimation(only linear distance now)
- Communication through mavlink protocol
- PID remote drone controlling
- Extra demands & functionalities emerge during development
Prerequisities:
- Python 3.x
- OpenCV python binding(with contribution library)
- MAVSDK python binding
To install them on a raspberry pi:
sudo apt install python3
pip3 install opencv-contrib-pythonMAVSDK installation: Python wheels of MAVSDK-Python on raspberry pi platform is currently not available. This gives extra complexity to deploying MAVSDK onto Raspberry Pi platform.
Basically, you'll have to clone the MAVSDK-Python repository from github and install the pip package manually. You'll also have to compile the MAVSDK backend(mavsdk_server process) from the MAVSDK C++ main repository and move it along side the Python binding libraries so that it can be invoked. Refer to Mavsdk-python github and cross compiling backend page from MAVSDK documentation to get step-by-step instructions.
Installing on windows follows similiar procedures. You'll have to pip3 uninstall opencv-python first if you already have the opencv python binding installed without contribution libraries, otherwise the two modules will conflict.
After cloning the repository, run aruco_detect.py to begin tag detection.
Run aruco_detect.py in a Python 3 interpreter to run the detection program. It uses cv2.VideoCapture(), so it should work fine on both laptop camerars and usb camera on a Raspberry Pi.