Fly around the World

Oshkosh Scenery

Visible light Landsat image of Lake Winnebago, WI

On occasion of the EAA AirVenture Oshkosh 2007 on July 23rd-29th the custom scenery project is preparing a custom scenery, capturing an area of about 6800 square kilometers around Lake Winnebago. The aim is to provide more accurate scenery for the area around Oshkosh, specifically those areas which are interesting for the special approach procedures.

The landcover data is derived by automatic landcover classification from Landsat imagery. The Oshkosh scenery represents up to now the largest area that has yet been processed with this method in the Custom Scenery project.

The scenery is available at

Issues encountered during classification

Unfortunately, Green Lake, which is used as a holding together with Flush Lake, is not completely covered by the landsat tile which was used.

The only available adjoining Landsat tile is 2 years younger (from 2001) than the other tile (from 1999) and has different lighting conditions (see image below), which makes it difficult to simply patch the two tiles together and use the same training information for classification.

A color-balancing method described by Wu and Campbell (PDF) may very well be a solution to that problem, but up to now no implementation of this in freely available software has been found. Therefore an own implementation, possibly within GRASS or more generally based on GDAL may be necessary.

Training data used for the Oshkosh, WI area (left),
Classification results (middle),
Comparison of adjoining landsat tiles at Green Lake, WI (right)

Use of TIGER line-data

The linear features in the Oshkosh scenery (roads, streams, railroads) will be based on the TIGER dataset by the U.S. Census Bureau. The data is pretty current and accurate, and in some points contains more data than TerraGear and FlightGear can actually handle.

The data used is therefore currently limited to highways, primary and secondary roads. Local roads are not included. In a next step, also other line data such as railroads will be included as well.