search

tacman-fp7 / handover

This module implements Task 4.4, to transfer an object from one hand to the other. It first localised the object in the robots hand. Then, the robot selects a pose for the second hand in which the object can be successfully transferred. The second hand grasps the object. Once a stable grasp is achieved, the first hand releases the object.
GNU General Public License v3.0
2 stars 1 forks source link

[iCubDarmstadt01] problems in running cameras #22

Closed giuliavezzani closed 7 years ago

giuliavezzani commented 7 years ago

Sometimes, when I try to launch cameras, one of them is not working (e.g. left). If I play with the yarpmotorgui changing the value of eye version of the left, I got some strange images on the left camera. The same happened also to the right.

But, sometimes they can be launched without any problems.

pattacini commented 7 years ago

@giuliavezzani it's likely then there's a problem with the firewire cable connecting the camera to PC104. Due to wear and tear and the fast movements of the bulb it tends to become loose.

giuliavezzani commented 7 years ago

thanks @pattacini. I'll leave here this issue just as a reminder for the next time, when technicians will be here again :smile:

lornat75 commented 7 years ago

I think the problem is in the camera flat cables, we can bring replacements next time, they are fairly easy to mount.

Sent from mobile device

-------- Original Message -------- Subject: [tacman-fp7/handover] [Reminder] problems in running cameras (#22) From: Giulia Vezzani To: tacman-fp7/handover CC: Subscribed

Sometimes, when I try to launch cameras, one of them is not working (e.g. left). If I play with the yarpmotorgui changing the value of eye version of the left, I got some strange images on the left camera. The same happened also to the right.

But, sometimes they can be launched without any problems.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://github.com/tacman-fp7/handover/issues/22, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ABVpxX_BPq4ceMzXeJ_H918ZKcWjLtrDks5rIll7gaJpZM4LO_sB.

giuliavezzani commented 7 years ago

Though this morning I managed to calibrate SFM and got quite good disparity map (see #17), now I can't calibrate SFM any more. I tried several scenarios: a lot of different objects on a table, a 3d scenario in the room, different SFM parameters, different cameras tuning etc.., but calibration never converges. In addition, when I rarely managed to calibrate SFM, the resulting point clouds are actually unrealistic and wrong.

lornat75 commented 7 years ago

Are there enough features in the images?

giuliavezzani commented 7 years ago

Yes, even when I have a lot of features.

However, the behaviour is not constant. Now, I manage to calibrate SFM, but the obtained point cloud are completely crazy.

Furthermore, I've also noticed:

eyes

I don't know if all these elements can have any influence on SFM, mostly since yesterday, in the same scenario, I got better results. I need to discover the source of such a lack of robustness..

giuliavezzani commented 7 years ago

Summary of all problems experience with cameras

  1. some problems in cables connection. Sometimes the cameras need to be moved a little bit to actually be connected and provide images

  2. the required cameras parameters for having similar image are much different between left and right

  3. problems in SFM calibration and reliability (even if, after all, I think I managed to find an acceptable calibration - at least for point cloud acquisition)