Tracking a 2D floor reference image on a mobile robot using an IP camera and returning the relative coordinates

[LTIERNEY]

Hi all,

I am in the process of making a vision system for a mobile robot that would detect a "goal square" on the ground and localise the robot to assist in it entering the square. The camera would be fixed and so the robot would just retain any final coordinates before the object is obscured.

I have been using the Vision Assistant for other applications like tracking a coloured cube's pixel coordinates but can't find an example to do this second task.

below is the raw square image i wish to detect from any approach angle, i have multiple of these but I'm only posting one as an environment example. I can isolate the tape from the background easily but I just need help with tracking it while it's on screen.

Thanks in advance.

[BruceAmmons · Accepted solution]

You should be able to isolate the tape from the background and get a binary image of the tape.  If necessary, clean it up so the largest object is the only object.

If you want, you could use IMAQ Convex and IMAQ Fill Holes to make it a solid object.  This might improve the results for partial boxes.

Use IMAQ Particle Analysis to find the properties of the object.  I think center of mass (X and Y) would probably be the best option.

This will give you a point you can steer toward, and will be forgiving if the entire box is not in view.  It will still give you the correct direction to move, but not necessarily the right distance.  After moving, you can repeat the process and keep correcting until it is centered properly.  I would turn left and right and get it centered before moving forward.

Bruce

[LTIERNEY]

I was already finding the binary target but I was having trouble figuring out what to use to get central coordinates for it.

You completely solved my problem, thank you ❤️