lardemua / atom

Calibration tools for multi-sensor, multi-modal robotic systems
GNU General Public License v3.0
253 stars 26 forks source link

Consider the detection of a charuco board #87

Closed miguelriemoliveira closed 4 years ago

eupedrosa commented 4 years ago

I got the charuco detection working.

charuco

The detection time is around 10ms, even if there is no board or only a partial part is detected. Here is an image of a partial detection. charuco_partial

Just to say that I already have a dataset for the eye-in-hand problem. I will try to work on this today. I have no idea where the camera is... IMG_20191211_112353

miguelriemoliveira commented 4 years ago

Hi @eupedrosa Looks good. What about the failures, i.e. when there is no charuco board for example, how does it behave?

eupedrosa commented 4 years ago

What about the failures, i.e. when there is no charuco board for example, how does it behave?

Behaves very well!!!! It is faster when there is no board. It is the opposite from the classic chessboard detector.

miguelriemoliveira commented 4 years ago

Sounds good, but to change all the code to arucos right now would be a gigantic task ... are you moving forward with the aruco hand in eye calibration? We, on our side, can just read the json and for now, if the marker is an aruco, abort sayin "unsuported marker type" ...

miguelriemoliveira commented 4 years ago

but what about the data collector and labeller ...

eupedrosa commented 4 years ago

Sounds good, but to change all the code to arucos right now would be a gigantic task ...

Maybe not. The charuco board also provides the corners of pattern. The information is the same but extracted with a different method. My idea is to support both patterns and atlas car dataset will work without a problem.

are you moving forward with the aruco hand in eye calibration?

Yes I am. Using charuco boards is a breeze. Stay tuned...