Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Some questions about reliability of identification #235

Open
meihe323 opened this issue Jun 16, 2022 · 3 comments
Open

Some questions about reliability of identification #235

meihe323 opened this issue Jun 16, 2022 · 3 comments

Comments

@meihe323
Copy link

Im working on a project that landing a copter on the apriltags ,here comes some diffculties,the first is that when the copter is very close to the apriltag, I found that the shadow of copter covered some black blocks of tag, and I can't get good identification result.How can I do to deal with shadows to get better reliability of identification?

@brmarkus
Copy link

Would you have a chance to add diffuse lights/LEDs to the bottom of the copter? Or add multiple tags next to each other? Or add a circle in a thicker line around it then you can switch the algorithm from Apriltag-detection to tracking the circle's center.

@christian-rauch
Copy link
Collaborator

Do you mean you get issues when the shadow covers the white blocks? If the contrast between the dark and light areas of the tag (you can use different colours, it's just about the edges) is too low, the edge detector will have issues. You can try to play around with the exposure value or manually threshold the image into dark and light areas and feed this image to the AprilTag detector.

@chutsu
Copy link

chutsu commented Oct 27, 2022

Very late to the party, but way back when I was doing a similar project; landing a quadcopter onto an AprilTag in outdoors. We basically used the "illumination invariant transform" proposed by Will Maddern:

Source:

Maddern, Will, et al. "Illumination invariant imaging: Applications in robust vision-based localisation, mapping and classification for autonomous vehicles." Proceedings of the Visual Place Recognition in Changing Environments Workshop, IEEE International Conference on Robotics and Automation (ICRA), Hong Kong, China. Vol. 2. No. 3. 2014.

Link: https://www.robots.ox.ac.uk/~mobile/Papers/2014ICRA_maddern.pdf

The only requirement is the camera has to be RGB, and by manipulating the RGB channels via the transform you can form a 1 channel illumination invariant image. The result is if you print a color AprilTag you can mitigate the affects of shadows outdoors like so:

Screenshot from 2022-10-27 12-07-59

https://www.youtube.com/watch?v=CgcqxZ-QGTc

Page 33 (Landing Target Detection): https://uwspace.uwaterloo.ca/bitstream/handle/10012/14192/Choi_Christopher.pdf?sequence=3&isAllowed=y

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

4 participants