Fisher Massif Mapped from Landsat 7 Imagery.

Brief description

Fisher Massif Features Mapped from Mosaiced Pan Sharpened Landsat 7 Imagery.

FEATURE MAPPING
An unsupervised classification was run on the final image to create an image with 12 distinct grey scale values. An automated feature extraction process was then performed in ERDAS to automatically select and extract areas of Rock and Snow. These areas were then compared with the true colour image mosaic and the boundaries were manually adjusted where necessary. All other feature types were mapped in ESRI's ArcGIS by manually tracing along and around features using a stream digitising technique. Relevant linear features were then polygonised.

The accuracy of the mapping was within +/- 30m for 95% of mapped features. This is a relative accuracy as there were no control points available to provide an absolute image orientation.

The datasets were converted to double precision ArcInfo Coverages in UTM Zone 42. ESRI?s ArcGIS Desktop and Workstation were used to process the vector data.

The Rock and Snow features which had been automatically extracted from the image were generalised using a distance of 40m with the bend simplify option of remove redundant vertices. The lines were then splined with a grain tolerance of 20m to smooth them.

The line feature coverages were cleaned with a tolerance of 0.1m. ArcEdit was then used to tidy the line work in the resultant coverage. Arcs were extended where required and overshoots were deleted. The resultant coverages were built for line and polygon topology.

The polygons were attributed using the Landsat image as a backdrop. The features were then extracted into separate feature coverages. The data were attributed according to the AADC Feature Type Catalogue. The individual feature coverages were built and checked for errors. A further visual check was then performed to check the features corresponded to the image.

The absolute accuracy of the features mapped is +/-280m, with a relative accuracy of +/-30m.

The individual feature coverages were projected to Geographicals (WGS84).

After discussions with Mike Verrier, from the AAD on 30 April 2003, it was decided that ridgelines would only be picked up where there were major variations in the surface and not where there was a small hollow in which snow was settling.