Adding the Sky Gradient to WWJ

[patmurris]

Following on my first try at a renderable layer for WWJ, i'm in the process of porting the Sky Gradient code too... but am having some kind of 'reference center' problem (not seen in the screenshot though).



The code renders a sphere portion centered on the eye (camera) so that it intersect the globe exactly at the horizon. It is constructed at the origine, rotated to account for the camera position latitude and longitude, and then translated toward the camera.

This code worked in WW and WW2DPlusOne and it works too in WWJ except the 'sky' sphere position seems to be 'shifted' by some amount and doesnt render in the right position. It gets better when the camera comes close to the ground.

I'm wrondering if there is some sort of 'floating reference center' that should be accounted for at some point ? tag ?

Note: i'm expecting some problems with the ellipsoidal globe since the sky geometry is a regular sphere here. I will deal with it after it falls in the right place first.

[fabrizio.giudici]

First of all, thank you for your posts! Using the previous one I was able to understand a lot about how to add "things" to the scenario.

I'm experiencing too some positioning problems. I'm exercising by writing a layer that draw parallels and meridians. I just modified your code and replaced the stars with a mesh. Now I managed to render a spherical grid, but:

1. the sphere is not centered with the Earth;
2. I had to use equatorialRadius() * 1.5 to roughly match the Earth radius

Now I'll try to attach a screenshot.

[fabrizio.giudici]

Here it is:

[patmurris]

I'm glad you could get started with the StarsLayer code
Does your grid position shift when the camera moves or tilt ?

[fabrizio.giudici]

Originally Posted by patmurris

I'm glad you could get started with the StarsLayer code
Does your grid position shift when the camera moves or tilt ?

I don't know. Indeed it was my very first impression, but then you made me discover that its shifted. So it could be a perspective thing. Now I'll try to manually recenter it and let's see.

[fabrizio.giudici]

I've found this method:

Point pos = dc.getGlobe().computePointFromPosition(Angle.fromDegreesLatitude(lat), Angle.fromDegreesLongitude(lon), h);

which centers correctly the stuff. Strangely enough, I had to set h = 3400000 to fit the Earth size - I expected 0 if it referred to the ASL, or 6349etc if to the center of the planet.

But there are problems still. Rotating the Earth, the meridians "slide" over the terrain ; if I tilt the view by means of shift+up/down, the grid completely detaches from the Earth

[patmurris]

Originally Posted by fabrizio.giudici

Rotating the Earth, the meridians "slide" over the terrain ; if I tilt the view by means of shift+up/down, the grid completely detaches from the Earth

thats what i was refering to by 'floating reference center'...
We need tag advise here...

[patmurris]

Check the Polyline class in geom. It draw GL lines according to points positions (lat/lon/elevation). Cartesian corrdinates are computed with

dc.getGlobe().computePointFromPosition(p.getLatitude(), 
    p.getLongitude(), p.getElevation());

It computes a local reference center that is substracted from all vertices coordinates when building the geometry and then is 'pushed' during rendering :

dc.getView().pushReferenceCenter(dc, this.referenceCenter);

[tag]

Please post your graticule code and we'll try to help.

In order to achieve precision over great distances, most geometry is calculated relative to a local origin. But code that doesn't need to worry about very fine precision doesn't need to concern itself with that. At least we thought we designed it that way.

A caution: A graticule is a lot more than lines around the globe. If you simply use lines some distance above the surface, the lines shift relative to the ground as you move. For me, the equator should touch the same ground no matter how I view it. Lines also pierce the terrain unless they're so far off the surface as to miss Mt. Everest, or they follow the terrain, which we implemented but decided was still unacceptable. We know what to do; we just haven't done it yet.

[patmurris]

Here is my actual version of the sky gradient (zip 3k).
I guess the way i compute cartesian coordinates and 'move' the geometry before drawing has something wrong. Maybe i dont use the right values from the current view... see comments in the source.

If you can have a look. Thanks.