Stereo Calibration on Zed X One tips

[adinesh]

Starting a thread for questions related to the stereo calibration setup, in case anyone runs into the same issues.

1. what does it mean to have a negative baseline at the end of the stereo calibration?

   1. Answer: it means your left and right cameras have been flipped. This is what led to the images looking ‘corrupted‘ when I tried to view them through ZED_Depth_Viewer.

2. have you tried working with configuration with very large baseline? Mine is 170cm. this means that the coverage we can achieve in step 1 stereo_calibration script is only ~60%

3. In the ZED_Depth_Viewer, the point cloud and the depth map are only displaying objects that are present in the left camera. However, it must be using both cameras somewhere because the metric estimation is not bad. Can I see the point cloud and depth map of the stitched frames?

   1. it is repeated here that retrieveMeasure() returns “depth data aligned with the left image“. can i align it with the right image?

4. Are there good docs anywhere on working with virtual cameras?

[Myzhar]

Hi @adinesh

@adinesh

what does it mean to have a negative baseline at the end of the stereo calibration?

Answer: it means your left and right cameras have been flipped. This is what led to the images looking ‘corrupted‘ when I tried to view them through ZED_Depth_Viewer.

This is correct. swapping camera connections normally solves this problem.

@adinesh

have you tried working with configuration with very large baseline? Mine is 170cm. this means that the coverage we can achieve in step 1 stereo_calibration script is only ~60%

The tables in this support page allow you to understand more on what you should expect when changing the baseline: https://support.stereolabs.com/hc/en-us/articles/20851232732311-Setting-up-your-Stereo-Rig-with-two-ZED-X-One-cameras

@adinesh

In the ZED_Depth_Viewer, the point cloud and the depth map are only displaying objects that are present in the left camera. However, it must be using both cameras somewhere because the metric estimation is not bad. Can I see the point cloud and depth map of the stitched frames?

Depth is mapped only to one camera. This is expected, and this is how stereo vision works.

The ZED SDK allows you to compute depth for the right camera, but we do not recommend to process both simultaneously because it would require double the computing power:
sl::InitParameters::enable_right_side_measure

@adinesh

Are there good docs anywhere on working with virtual cameras?

Virtual stereo cameras are used in the same way you use standard stereo cameras.
Only the setup phase changes, because you must calibrate the system and use the ZED Media Server: Setting up the ZED X One Stereo - Stereolabs