Techniques to Improve Elevation Rendering Time

[bgien1]

Hello,

I'm currently using an application that uses osgearth 2.10.1, and I'm attempting to create a map, with terrain, for offline use. Ideally I'd update to a newer version of osgearth, but the application I'm using is developed by a separate organization.

Below is my .earth file, as you can see I am loading in a .vrt (which points to my dted, which I've made sure only contains .dt1 files), as well as a raster .mbtiles. My issue is that in my application at runtime, the rendering process is extremely slow. I believe this is related to building the heightfield, as when I use the map with raster tiles only (no elevation block), it is quite fast.

I've also tried using a .vrt which is based off of a DTED0, which is significantly smaller, but this also presents the same rendering slowdown issue. My question: Is there some best practice that I'm missing here? Is there a more efficient format for storing and querying elevation, and tiling it in osgearth?

<map name="my elevation">
    <!--Load a folder full of terrain data as an elevation source-->
    <elevation name="my_elevation" driver="gdal">
        <url>./worldwide_dted1.vrt</url>
    </elevation>

    <image name="my medium res map" driver="mbtiles">
        <url>./satellite_2016_med_res/maptiler-satellite-2017-11-02-v1.2-z0-z13.mbtiles</url>
        <compute_levels>false</compute_levels>
    </image>
</map>

[gwaldron]

Not sure if it will work in 2.10, but you can try using osgearth_conv to build an mbtiles elevation dataset.
Something like

osgearth_conv --in driver gdal --in url something.vrt
              --out driver mbtiles --out filename myData.mbtiles
              --out format tif --elevation --threads 8

That's off the cuff; things were different back in 2.10 :)

And don't forget that DTED data is relative to the EGM96 ellipsoid, so you should tell osgEarth about that in your layer.

<vdatum>egm96</vdatum>

Failing that, you can try using a tool like QGIS or Global Mapper to export a tiled dataset. Good luck.