cam_test camera_center_diff

[jlaura]

The cam_test documentation states:

It find(s) the camera center and ray direction at a set of sampled pixels for both cameras and checks how they differ. (typo in 'find').

Are the reported values the difference in the sensor location in x,y,z space w.r.t. some datum or the signed(?) deltas w.r.t. the cam1 location?

[oleg-alexandrov]

Jay, Thank you. I fixed the typo. I also expanded the doc a bit: It now says: This tool compares two camera models for the same image, or a camera model against itself. It finds the camera center and ray direction at a set of sampled pixels for both cameras and checks how they differ. Then, it projects pixels to the ground using the first camera and back-projects the resulting ECEF points into the second camera, and estimates the pixel discrepancy. So, three comparisons are done (1) camera center, this is in units of meter (2) ray direction, (3) projection into camera, with the latter in units of pixel. It does not compute signed deltas, only norm of of those 3D and 2D vectors (for camera centers and pixels, respectively). Let me know if there's a need for more detail. I did not include a sample output for this tool in the doc, as being too verbose, but for the record, it is the following, and it is also suggestive about what the tool does: cam1 to cam2 camera direction diff norm Min: 8.67632e-10 Median: 1.7293e-07 Max: 9.0646e-07 cam1 to cam2 camera center diff (meters) Min: 3.79606e-06 Median: 5.1208e-05 Max: 0.308461 cam1 to cam2 pixel diff Min: 4.03812e-05 Median: 0.0122446 Max: 0.0465756 cam2 to cam1 pixel diff Min: 4.83225e-05 Median: 0.0121175 Max: 0.0466092

…

On Mon, Dec 5, 2022 at 7:43 AM jlaura ***@***.***> wrote: The cam_test documentation <https://stereopipeline.readthedocs.io/en/latest/tools/cam_test.html> states: It find(s) the camera center and ray direction at a set of sampled pixels for both cameras and checks how they differ. (typo in 'find'). Are the reported values the difference in the sensor location in x,y,z space w.r.t. some datum or the signed(?) deltas w.r.t. the cam1 location? — Reply to this email directly, view it on GitHub <#388>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAKDU3HPLHDU7Y5UQHVQDDLWLYER7ANCNFSM6AAAAAASUNSVVU> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[jlaura]

@oleg-alexandrov That's terrific. Thanks! I have been using this also as an easy method to see what the adjustments look like in sensor space for images data pre- and post-adjustment. So, a side benefit because the tool is so convenient to use! Many thanks for developing and releasing it.

[oleg-alexandrov]

And I am not sure if I actually answered your questions. All xyz measurements of sensor location are in ECEF, which kind of seems like the logical coordinate system for this tool as those are orbital data rather than on planet surface. On Mon, Dec 5, 2022 at 10:19 AM Oleg Alexandrov ***@***.***> wrote:

…

Jay, Thank you. I fixed the typo. I also expanded the doc a bit: It now says: This tool compares two camera models for the same image, or a camera model against itself. It finds the camera center and ray direction at a set of sampled pixels for both cameras and checks how they differ. Then, it projects pixels to the ground using the first camera and back-projects the resulting ECEF points into the second camera, and estimates the pixel discrepancy. So, three comparisons are done (1) camera center, this is in units of meter (2) ray direction, (3) projection into camera, with the latter in units of pixel. It does not compute signed deltas, only norm of of those 3D and 2D vectors (for camera centers and pixels, respectively). Let me know if there's a need for more detail. I did not include a sample output for this tool in the doc, as being too verbose, but for the record, it is the following, and it is also suggestive about what the tool does: cam1 to cam2 camera direction diff norm Min: 8.67632e-10 Median: 1.7293e-07 Max: 9.0646e-07 cam1 to cam2 camera center diff (meters) Min: 3.79606e-06 Median: 5.1208e-05 Max: 0.308461 cam1 to cam2 pixel diff Min: 4.03812e-05 Median: 0.0122446 Max: 0.0465756 cam2 to cam1 pixel diff Min: 4.83225e-05 Median: 0.0121175 Max: 0.0466092 On Mon, Dec 5, 2022 at 7:43 AM jlaura ***@***.***> wrote: > The cam_test documentation > <https://stereopipeline.readthedocs.io/en/latest/tools/cam_test.html> > states: > > It find(s) the camera center and ray direction at a set of sampled pixels > for both cameras and checks how they differ. (typo in 'find'). > > Are the reported values the difference in the sensor location in x,y,z > space w.r.t. some datum or the signed(?) deltas w.r.t. the cam1 location? > > — > Reply to this email directly, view it on GitHub > <#388>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AAKDU3HPLHDU7Y5UQHVQDDLWLYER7ANCNFSM6AAAAAASUNSVVU> > . > You are receiving this because you are subscribed to this thread.Message > ID: ***@***.***> >

[jlaura]

@oleg-alexandrov Both of these are terrific answers. I'll make the first one, the second one answers my question as well.