Hi all, I’m currently working on computer vision. After doing my research, I have found two methods for estimating the camera’s position and orientation using just one camera.
The first method involves using optical flow to estimate the x and y directions with a circle detection to get the distance from that object (z axis)
the second involves using Aruco and OpenCv .
has anyone tried one of these methods with pixhawk
Yes, lots of people do that successfully. And most of them use ArduCopter 4.4.4 or newer.
Thank you for confirming! I will dig deeper into exploring these techniques and their integration with Pixhawk for my project. If you have any tips or resources you could share to get started, I would greatly appreciate it. Thanks again for your help!
For loiter indoor, I have used Hereflow + down facing rangefinder, pilot flying. If you bring in other middleware then you are probably talking about indoor autonomous flying.
I pause the work with Arducopter after trying to reuse the legacy camera module failed to interface well with opencv 4.x. my exploration is using ROS 2 Humble with Ubuntu 22.04, tier 1 support, direct installation. I believe no gps guide mode is needed.
I am just lazy to go through the calibration of the camera hoping to reuse the calibrated data.
This group development worked on PX4 stack with Pi4B. Mavros 1.
I have never worked with ROS, and I know it would take too long to learn. I’m simply using ArUco markers with DroneKit. Could you please provide me with the versions of DroneKit and Python you are using?
Thank you for your help. I’ve already tried it, and it worked. However, I’ve
faced compatibility issues with OpenCV and Python 2.7.16. I’ll explore MAVSDK as an alternative. Could you please provide me with a link where I can get started?
Already in the link. Have you use GitHub before, check out the readme?
By the way, if you intend to use openCV and with ArduPilot, then you should use Mavros 2 (mavlink + ROS). Then VIO tracking camera for non-GPS navigation may be of your interest.