Items with no label
3338 讨论

Scale problem in position

SMish7
新手
2,340 次查看

Hi,

I think there is some sort of scale problem in the positions given by realsense slam on the Euclid. Is there any parameter available to tune and make the scale correct. For eg. the depth map factor that is present in the ORB_SLAM2.

The attached picture conveys what I mean here:

The ground truth is given by the motion capture system.

0 项奖励
4 回复数
Meitav_K_Intel
1,100 次查看

Hi,

From my experience, if the depth camera in the device loses calibration, it tends to result in a change of scale.

Can you try making some simple depth measurements to see if you're getting any scale error in them too?

As far as I know, the realsense slam doesn't have something like the depth map factor that ORB_SLAM2, but if my suspicions are true, recalibrating the device should solve it.

Regards,

Meitav

Intel Euclid Development Team

0 项奖励
SMish7
新手
1,100 次查看

Hi,

I am a novice here. Could you please tell me how I can do simple depth measurements and also how can I calibrate the device? I can calibrate the fisheye and the color cameras but how can I calibrate the IR cameras?

Regards,

S.

0 项奖励
Meitav_K_Intel
1,100 次查看

Hi,

We are planning on making a ROS node for recording images in real time. In the meantime it might be a bit advanced to implement saving on one of the realsense samples provided with librealsense' source.

As for calibration for the IR+Depth, we have a tutorial for that here: https://click.intel.com/media/productid2100_10052017/dynamic_calibration_guide.pdf https://click.intel.com/media/productid2100_10052017/dynamic_calibration_guide.pdf

Please let us know if this gives you better results.

Regards,

Meitav

Intel Euclid Development Team

0 项奖励
idata
员工
1,100 次查看

Hello SubMarIn,

 

 

I was wondering if you require further assistance.

 

 

If you have another question, don't hesitate to contact us.

 

 

I'll be waiting for your response

 

 

Have a nice day.

 

 

Regards,

 

Andres V.
0 项奖励
回复