Christoph Kaeser
3 INPUT DATA
The major input data is VECTOR25. VECTORGS is generated, using as background the pixel-maps (scanned national
maps), with the following software: the semi-automated line-following software VTRAK from LASERSCAN for
transportation network and forest limits ; MicroStation from BENTLEY for manual digitisation of areas and single
symbols. For automatic extraction of buildings, L+T plans to use the KAMU/AUTOVEC pattern recognition and
vectoring software developed at the ETH Zürich (Stengele, 1995; Frischknecht and Kanani, 1997). The transportation
network exists over all Switzerland in vector format since the end of 1999. VECTOR25 have an RMS error of ca. 2.5-
7.5 m and a maximum one of ca. 12.5 m (based on empirical values), including generalisation. They are topologically
correct, but due to their partly automated extraction, some errors might exist. In addition, errors in orientation, size and
shape of the objects, esp. the buildings, exist mainly due to generalisation. Figures 1 and 2 show some VECTOR25 data
overlaid on the aerial images. Regarding buildings, occurring differences between VECTOR25 and reality include
shifts, rotations, more objects in one VECTOR?2S5 buildings, digitisation errors (see top right image) etc. Regarding
roads, problems occur in forests and their boundaries, urban areas (occlusions, shadows, trees, cars etc.). The general
shape of VECTOR25 is correct (with more problems at intersections and squares), but even in this case vectors may
pass on top of buildings, trees etc. Even the spectral properties of the roads may vary (see left and right image on
bottom row of Fig. 2). For ATOMI, the VECTOR25 data are delivered in DXF format as polylines, together with any
existing class attributes. The objects of VECTOR25 data are digitised for each map sheet into different DXF-layers in a
so-called geometric-based form. To exploit the full potential of this vector data over Switzerland, they have to be stored
in a GIS. This was the main reason for the establishment of a Geo-Topographical DataBase (GTDB). The system used
is ARC/Info with Spatial Data Engine from ESRI using an ORACLE database running under Unix on an IBM platform.
VECTOR25 consists at the moment of more than 3 million objects. In GTDB, they are integrated into the so-called
semantic model. This model allows searching, e.g. for villages or rivers, and all the related objects are selected and
displayed.
Figure 1. Examples of VECTOR?2S data for buildings.
Other input data are: the nation-wide DTM (DHM25) with 25 m grid spacing and an accuracy of 2.5 m in the lowlands
and 10 m in the Alps ; the raster map (PIXELMAP) with its 6 different colour layers and the digital images. Other
derived data used in the project, are DSMs (1-2 m grid, from 14-28 um images) and colour orthoimages (0.25 m pixel
size), generated with PHODIS OP and TS. Currently, a test dataset in a region with representative topography and
landcover for Switzerland is used (Albis, close to Zurich). The colour imagery has scale 1:15,000 (30 cm focal length),
25% sidelap, and was scanned with 14 microns at a Zeiss SCAI scanner. At this region ground truth data (DTM, road
464 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.