Camera mounting positions

PaulKeeling

Hi, 

I was wondering how does the camera handle different mounting positions?

Cheers, 
Paul

Grady_Turner

@PaulKeeling  I assume you mean mounted so that the locating hole is not lined up with the one on the flange?

My guess is that the offsets (relative moves) under the camera locate node will all be wrong because the axes being commanded do not line up physically with the way the camera is mounted.  Same as the force torque sensor.  If it is mounted incorrectly it will report forces along the wrong axes.

Alexandre_Pare

@PaulKeeling @Grady_Turner
I did not give this a try yet. From a mathematical point of view and based on vision knowledge, during the calibration process of the camera, the camera is taking several snap shots to find the vector relation between the camera position and the calibration grid. So at the end of these mathematical calculations you would end up with the correct vector relation that would define the camera TCP to which all pictures of the parts would relate. So it could potentially work. However, the automatic position that the URCaps ask the robot to move to during calibration are based on the camera mounting position that we have designed for. So this may be an issue to run this with the camera mounted in a different way since the positions may not be adequate for you to do a correct scan during the calibration procedure. @JeanPhilippe_Jobin can you check this question for further details on this?

JeanPhilippe_Jobin

Never been tested but it should work (under certain limits). The calibration uses the suggested mounting position as an initial position for the calibation but nothing is specifically hardcoded to this specific mounting position. But this would have to be tested. 

Grady_Turner

@JeanPhilippe_Jobin @Alexandre_Pare  This answer does not make sense to me.  In the calibration and part teaching a Feature point is programmed within the robot and is relative to the coordinates of the TCP in the base coordinate system.  In the program all the offsets are done using this feature.  I would think that the if the camera were rotated on the mount flange, then all of the coordinates would be off by that rotation.  Is this incorrect?

If so, does the calibration routine "notice" that rotation offset and account for it in the program?

LouisAlexis_Demers

Good question, I think that if the calibration process converge to a solution, then it should work afterward.

Anyway, @PaulKeeling, we strongly ask to mount the camera in the correct orientation on the robot flange. This is why the dowel pin is press-fit into the camera. Hence, you are sure that no matter whom will take off the camera from the robot, the camera will always be in the same orientation when it will be replaced.

Alexandre_Pare

@Grady_Turner
I will try to explain this better. What you are doing during the camera calibration procedure with the grid is finding out where the object is relative to the robot base frame. So the algorithm in that automated camera calibration procedure figures out where the camera actually is on the robot to be able to translate the object position from the camera coordinate frame to the robot base coordinate frame. This is all done using linear algebra concepts. Once that relation is found, the rest of the program is the same no matter what calibration value was found during the calibration procedure. So the mathematical concept behind that calibration has the potential to be able to figure out the object position relative to the robot base frame no matter how the camera is installed on the robot. However, Robotiq has not tested other mounting position so does not support this. The camera calibration algorithm has been optimized for the specific mounting position as shown on the picture below. Does this make more sense?

Grady_Turner

@Alexandre_Pare yes thanks, answers my question of whether or not the calibration routine takes care of mounting offsets