Water culling is too aggressive

[Amerio.Stephane]

Hi,

For "normal" view frustum, water culling is currently 'good enough'. But for our long range/short FOV view sensors, we need to have a very narrow FOV and a very long view distance (>20km for FOV<<0.1°) and the water culling is "a bit" too aggressive as it can produce images with missing water tiles, or even almost no water in it:

 

Yes, there should be water everywhere, it's not the horizon or the viewing distance clipping it here. Granted, a very small view position or angle change may restore the rest of the water, but still this will produce "flashing" images that throws off the sensor end user.

(Here is the setting for the above image: water settings low, far=100k, camera at 6850.67021, -13347.47211, 635.73370; angles 73.68980, 64.91969, 14.84274, FOV 0.1, far=100k. Just moving a tiny bit can change the culling.)

I tried all water geometry preset (low, medium, high, etc), with or without the "culling aggressive" checkbox, manually tweaking custom parameters, there is always some situation where the water is clipped too much.

[bmyagkov]

Hello Stéphane,

Thank you for bringing this to our attention. We were able to reproduce the behavior on our side quite easily using the configuration you provided.

With the “Low” quality preset this result can be expected to some extent as its parameters are rather limited for scenarios involving a very narrow camera angle. Even the “Ultra” preset offers only a slight improvement and it seems that the “Custom” preset also reaches certain engine-side limits.

Still setting the "Geometry Preset" to custom and increasing the water "Visibility Distance" value should provide the best results among the currently available options. We also suggest adjusting the Geometry Progression and Geometry Polygon Size values as these have shown the most noticeable influence on this behavior when tuned together.

We’ve already created a dedicated ticket in our internal tracker to explore possible improvements on the engine side hoping this aspect can be enhanced in future releases.

Thank you once again for your detailed report.

[Amerio.Stephane]

Hello,

Do you have some information on the evolution of this ticket? This one is becoming problematic for long distance / high zoom SAR evaluations. Any workaround?

[bmyagkov]

Hello Stéphane,

42 minutes ago, Amerio.Stephane said:

Do you have some information on the evolution of this ticket? This one is becoming problematic for long distance / high zoom SAR evaluations. Any workaround?

This specific issue is planned to be evaluated during the stabilization stage before the next SDK 2.21 update with the hope that it may be included in the release.

Thanks!