Detecting markers

asked 2017-03-20 06:59:01 -0500

dgrixti gravatar image


I wrote a python script which uses in order to detect AR markers on a wall and move the robot towards it.

I tested this in Gazebo simulation and it works but now I want to do this on a real robot and I need to detect a particular kind of marker illustrated below. When I try the alvar library and place this marker in front of the robot (with Kinect) it detects many marker points which aren't even there so it's a bit confusing.

image description

I am not sure if an AR can be generated in this format, it is more of a QR if I am not mistaken so could this be the problem?

I was wondering if there was a library which takes an image in PNG format and then detects the image accordingly giving me the pose and depth of the marker from the robot. I saw that the Alvar library can search for objects in some sort of XML format but I did not understand well if it's possible to generate a marker in a specific shape like in my case.

I also saw this library which could be promising but it lacks detailed documentation unfortunately so I am not sure which is the better solution.

I would appreciate any tips you can have on this subject.

edit retag flag offensive close merge delete