Aerial Imagery in LumenRT

I am working on a LumenRT model for a medium sized civil project. I have exported my existing surface to a Descartes Scalable Terrain Model and successfully draped an aerial image on it. However, when I export to LumenRT I get weird results. 

If I simply export to LumenRT (normal model) the aerial image on the surface looks terrific, but the civil elements look terrible. Alternatively, when I export to a "large" LumenRT model, I get civil elements that look great, but the aerial image does not come in.

What can i do to make the aerial image look as good in the "large" LumenRT model as it does in the regular LumenRT model?

I am using LumenRT Version 16.01.30.66 and Microstation SS4 Version 08.11.09.832

Thanks,

-Ben

Attachments:

Regular-1.jpg: "Regular" LumenRT Model

Regular-2.jpg: "Regular" LumenRT Model

Large-1.jpg: "Large" LumenRT Model

Parents Reply
  • Hi Benjamin,

    The full resolution aerial images are not taken across from LumenRT. Here is what happens depending on the type of terrain used:

    1. If using a MicroStation Terrain Model, the aerial maps are composited into a 1K x 1K image (V8i) or 4K x 4K image (Microstation Connect Edition)
    2. If using a Descartes STM (this is your case), the images are broken up into 512 x 512 or 1024 x 1024 tiles depending on the resolution of the terrain – moving closer to the terrain with the camera will generate larger image tiles.
    3. If using a RealityMesh for the terrain, these are tiled at 512 x 512

    I hope this explains

    regards
    Mara
Children
  • Hi Mara,

    What's not clear from your explanation is the actual size of the tiles being compared.

    If I had a 1km x 1km draped mesh, with the draped image being 8000pix x 8000pix, are you saying that V8i would resample the single image down to 1000pix x 1000pix - or create 8 tiles of 1000pix x 1000pix each?

    Similarly, would CE more efficiently create two tiles of 4000pix x 4000pix?

    Essentially, which is the best method to create the highest resolution? From your notes it looks like;

    Best = Connect Edition draped mesh
    Next best = V8i mesh draped mesh
    No too good = Descartes STM & Reality mesh.

    Also, what is the purpose of "large model LumenRT export" relative to the standard export? Does it control resolution output, or fidelity of imported geometry?

    Max
  • Mara,

    I am still somewhat confused.

    Why does small model export versus large model export handle the exact same draped image differently? Why is the small model export properly shown with bright colored aerial image pixels but the large model export black? It looks like a software bug.

    Are you suggesting that if I split up a high resolution image into a bunch of smaller image tiles, LumenRT will successfully show the drape at full resolution in either large or small model export?

    -Ben
  • Hi Max,

    The downsampling of images draped on meshes during export is a function of the type of terrain mesh being used and whether or not you are using V8i or Connect.

    Let's say I had aerial imagery consisting of 8 images I draped onto the terrain:

    1). For standard Terrain Models (TM) found in basic MS and things like InRoads, even if you used multiple 8k x 8k images draped on the terrain, the output to LumenRT would be downsampled to s ingle 1k x 1k (V8i) or a single 4k x 4k image (MS Connect)

    2). For Descartes Scaleable Terrain Models (STM), the 8 draped images would be turned into mesh tiles depending on how close or far away from the camera the STM is. At a medium range distance, you would get something like 8 tiles each having 512 x 512 resolution, so this would be higher effective resolution than regular Terrain Model (TM) output.

    3). The highest possible resolution would be the new Reality Meshes. These basically output 512 x 512 or 1024 x 1024 images for all the tiles with each mesh tile being quite small. Also, in LumenRT Update 2 (later this month) you will be able to use the scaleable reality meshes meaning that there is no fixed resolution -- it will use all of the input maps resolution depending on how close the camera is to the mesh.
  • LumenRT Large Model Export is a V8i-only export facility to accommodate 32-bit operation on large models to prevent crashing. It basically does a 2 stage export so it's slower but more reliable than standard export. It has no bearing on the output resolution of image/aerial maps.