Anselm Schmieder 
  
AUTOMATIC DERIVATION OF GENERALIZED CONTOUR LINES FOR TOPOGRAPHIC MAPS 
USING HIGH-RESOLUTION AIRBORNE INTERFEROMETRIC RADAR DATA 
Anselm SCHMIEDER and Reinhold HUBER 
Aero-Sensing Radarsysteme GmbH 
c/o DLR Oberpfaffenhofen 
82234 WeBling 
Germany 
anselm C hans.as.op.dlr.de 
KEY WORDS: Interferometric SAR, DEM, cartography, contour lines. 
ABSTRACT 
Derivation of contour lines from a Digital Elevation Model (DEM) has to cope with a number of problems. (1) bad 
approximation of angular lines, (2) none or badly placed line inscriptions, (3) discontinuity when combining contour line 
sets of adjacent map sheets. We present a software system capable of dealing with the mentioned problems. A Raster- 
DEM derived from interferometric synthetic aperture radar (InSAR) AeS-1 data, operated by AeroSensing Radarsysteme 
GmbH, serves as input for the software system. Prior to vector processing of the contour lines, DEM smoothing and local 
peak detection are perfomed on the raster data. Contour lines are organized in point and polygon lists. Various algorithms 
operate on the polygon/point list data structure, e.g. breaking of contours touching the map border, removal of small 
contours by thresholding on the number of tie points. The core of the program is the generalization module performing 
line smoothing and generalization. Three different approaches have been implemented: take each Nth point, Pavlidis and 
the Douglas-Peuker algorithm. The latter performed best with respect to contour line quality / time consumption tradeoff. 
To ensure continuity between map sheets, an overlap area forming a skirt on all four borders is constructed and contour 
lines are cutted in appropriate fashion. Another important capability is automatic generation of line inscription. Graphic 
attributes for lines and inscriptions, especially font size, line type and thickness, inscription insertion point and angle are 
derived from the map scale requirements and local line properties. The resulting DXF-file consists of three layers: the 
main contour lines, the intermediate contour lines and the text, i.e. the line inscription. Further use of resulting DXF-data 
in a GIS system, in our case ARCView, is possible. 
1 INTRODUCTION 
Contour lines are the main graphical element to characterize 3-dimensional terrain on 2-dimensional map sheets. In this 
paper, an automatic way to derive terrain characterization employing contour lines is investigated. Automation of contour 
line extraction completes automated geocoding, georeferencing, classification and object detection from the remote sens- 
ing product. Contour line extraction from a DEM usually results in a large number of contours. Each contour is described 
by a polygon, which in turn consists of a number of points connected by straight lines. Simplification of these polygons 
is aspired from two points of view: computational and cartographic. From the computational point of view, a smaller 
number of points, up to our experience a portion of 1/20, reduces storage and processing costs significantly. Concerning 
the second point of view, the cartographer attempts to maintain the character and overall impression of the contour, while 
selecting representative points and deleting the rest (Douglas and Peuker, 1973). 
In this paper we will not try to suggest new contour extraction or simplification methods, we will rather address the 
fully automatic derivation of the DEM and the subsequent, also performed automatically, derivation of topographic map 
sheets containing simplified contour lines and inscriptions. Embedding this system into a SAR/InSAR processing chain 
and combining it with automatic classification tools, a raw map product for further manual postprocessing in common 
GIS software, such as ARC View GIS, is provided. Results concerning time consumption, accuracy of contours when 
compared to ancillary data, and completed products are presented. 
All presented results are based on data provided by the high-resolution airborne AeS-1 interferometric X-Band SAR 
system (Moreira, 1996). 
2 GENERATION OF DIGITAL ELEVATION MODELS FROM INTERFEROMETRIC SAR 
SAR is an active sensor, which means that it illuminates the imaged area by transmitting and receiving microwaves. 
Interferometric SAR (InSAR) employs two receiving antennas as shown in Figure 1 (b). Whereas in the conventional 
  
272 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part Bl. Amsterdam 2000. 
  
CO