fixed mIsOmnidirectional in WbObjectDetection

[FrederikBark]

Description
Changes the calculation of mIsOmniDirectional to horizontalFov > pi / 2.

Related Issues
This pull-request fixes issue #6395

Additional context
Details on how the spherical projections were wrong before can be seen in #6395

[omichel]

Thank you!
May I kindly ask you to add a line in the changelog under "Bug fixes"?

[FrederikBark]

Does this work for you?
f650732

[omichel]

It looks good to me, but I believe this test could be also simplified now.

[FrederikBark]

That seems to be the case. I adapted it and used M_PI_2 instead of M_PI / 2.0 in the initial change now to remove the additional computation and stick to the use of M_PI_2.

[omichel]

That looks very good to me now.
Thank you!

[omichel]

@TheMangalex: This PR apparently broke the camera_recognition test, see: https://github.com/cyberbotics/webots/actions/runs/6386382958/job/17333860294#step:7:1254
May I kindly ask you to look into this and check the camera_recognition controller to understand why the test fails and eventually fix it?

[FrederikBark]

I'll try to look into it in the following days

[FrederikBark]

It seems that this broke the test for cylindrical projections which seem to be handled similar to spherical projections in this case..

Before PR:

After PR:

It seems to be a different issue which was created by this change, as the issue of close objects also existed for cylindrical cameras:

I'll have a deeper look, what causes this

[omichel]

Thank you for investigating this.

[FrederikBark]

I created #6403 as a draft which seems to fix the test and #6395
In this PR I also added another object to the recognition test which is perpendicular to the camera sensor plane and only partially seen, as this was a problem in #6395

I want to add another test which specifically checks for the error which was fixed in #6395 , so I'll come back to this tomorrow.