VTP Software Tutorials

Tutorial 3: Imagery

This builds on the previous Tutorial 2, so you are recommended to do that one first.

You may have noticed that in Tutorial 2, we didn't use satellite or aerial imagery.  Instead, we just colored the elevation by height.  Why not use imagery?

Starting the search for imagery in the USA, you know from Free Image Sources that our options will probably be:

In the case of Chattanooga, Tennessee, all four of these are in fact available.

Step 1: LandSat

There are several ways to download LandSat (and the other USGS imagery), one of the easiest being to use the USGS National Map website.  Using its web interface, one can download the following LandSat image, which suffers from low-contrast and non-true colors:

 

To try this image yourself (without learning the National Map site), click the picture above and right-click, say Save Link As (or Save Target As in IE).  Save it to the folder called GeoSpecific on your Data path.  The file name is chatt1_1024.jpg

In the Enviro Terrain Properties, select the image in the Texture settings:

Upon draping it on the terrain in Enviro, it's easy to notice that besides having unrealistic colors, the low resolution of the imagery (~ 30m) is a poor fit for the 10m DEM, looking blurry:

Step 2: Better LandSat

The poor colors can be addressed by getting the full LandSat RGB directly from a source such as the ESDI and assembling them yourself.  If you follow those steps and merge the LandSat Red, Green, Blue bands, you'll get something like this:

Again you can save yourself the steps by right-clicking the above image and saving the target to your GeoSpecific folder.  Set it as your terrain texture in Enviro and it looks like this:

You can see that although the colors are now more natural, contrast is so high that the city areas look white and the vegetation such a dark green it's almost black.

A third option is to do a pan merge step with LandSat, combining the 15m greyscale band with the 28.5m color bands.  Several high-end image processing software can do that step.  Here is the result of pan-merge using the inexpensive PixelSense LS program:

The city area are now much better, very clear and defined, but everywhere is dark with murky colors.  You can partially correct for this in an image program like PhotoShop, although different aspects of the image such as color noise will be brought out.

Step 3: LandSat from NASA WMS

NASA very conveniently provides LandSat which has already been pan-merged, via a WMS (web map server).  You can use VTBuilder to get this imagery directly.  The only drawback is that NASA serves the data only in Geographic Coordinate System (GCS, aka "lat-lon") but our model of Chattanooga is in a Projected Coordinate System (PCS), namely UTM.  NASA's WMS won't reproject the imagery for us, so we have to do it ourselves.

In VTBuilder, load your elevation (chattanooga2.bt) layer.  Menu command Layer: Convert Projection, and set it to "Geographic" and WGS84:

Press OK and VTBuilder will reproject the data.  Menu command Area Tool: Set to Full Extents.  Notice how the data is looks much wider and shifted.  A rectangle in UTM is not a rectangle in GCS, and vice versa:

Menu Area Tool: Request Image from WMS.  (This does require you to be online.)  NASA JPL's server is the default.  Press Query Layers and you'll see a list of layers.  The one we want is the first one, global_mosaic.  Set Format to GeoTIFF, send the result to a file on your disk, and press OK.

Now you'll need to reproject the GeoTIFF image to our projection (UTM zone 16 NAD83).  VTBuilder doesn't reproject images, so you should use any geospatial software available to you.  If you are a command-line person, the GDAL command-line application are included with the VTP software, so you can use them.  Personally, i use Global Mapper.  After reprojecting, crop the image to fit the elevation.  You may recall the origin and size:

The resulting image looks like this:

Again, the colors aren't very realistic.  The whole image is dark, urban areas are purple.  (There are other styles available on NASA's WMS, including one called visual, in which the buildings are not purple but the surrounding terrain is a strange shade of blue.)  All these problems can be fixed in PhotoShop or your choice of image program.  I expanded the dynamic range, selected all the purple areas and desaturated them, resulting in this image:

Download it (Save Link/Target As..) and set it as your terrain texture, and it should look fairly nice in Enviro:

That's pretty good!  For further study, you can try using the NASA WMS, ESDI, and your choice of image software to produce a better satellite image.  You could also try working with images of size 2k instead of 1k, to preserve a little bit more of the pan-merge 15m data.

From these simple examples, you can begin to appreciate that developing realistic imagery from multi-band satellite imagery is a complex process, both a science and an art.

Step 4: USGS High-resolution Orthoimagery

The USGS is acquiring "high resolution color orthoimagery for 133 most populated metropolitan areas of the United States."  It's also available from their National Map website.  Here's an example of that imagery in the city of Chattanooga:

The resolution and color accuracy are good... in fact, the resolution is too good.  At .3m pixels (around 1 foot) the data is really huge, and unfortunately the National Map requires you to download it at full resolution.  If you want, for example in this case, 10m imagery, you must download 0.3m data (900x more data) and do the downsample yourself.

To get a sense of how this very high resolution data looks in a 3D scene, download chatt_downtown_jpg.zip (340k) and unzip it to your GeoSpecific folder.  This is a 1024*1024 area from the National Map which was downsampled from .3m to 3m (making the data 100x smaller).  To make it even smaller for this tutorial, I converted it from a GeoTIFF to a JPEG with World file (.jpw).

In Enviro, set this file as an Image Layer:

Now you should be able to see downtown Chattanooga in higher detail:

18