AUTOMATIC CONTOUR LABELING
FOR SCANNED TOPOGRAPHIC MAPS
Chao-hsiung Wu
Surveying Engineering Dept.
University of Maine
Orono, ME 04469, U.S.A.
William Mackaness
NCGIA
333 Boardman Hall
University of Maine
Orono, ME 04469, U.S.A.
Commission 4
ABSTRACT:
In many mapping organizations, existing topographic paper maps comprise the most complete source of elevation data; the
objective of this research is to capture the elevation data (which is in contour form) in a GIS. The topographic relief plate which
shows no height information can be scanned into a GIS. The remaining task is then to label these isolines.
This paper discusses a methodology for the automatic labeling of the scanned image based on the separate input of a minimum set
of height information. The implementation was divided into two parts. First, an undirected contour tree was generated from
analysis of a scanned image. Secondly, a set of heuristic topological rules were applied in order to orientate the tree and thus
determine a minimum set of height information required to label all the contours. Problems in this partial implementation and
worst case situations are discussed.
Key Words: Cartographic, Raster, Spatial, Directed Graphs, Contour Labeling
INTRODUCTION
The generation of digital terrain models (DTM) for geographic
database is an important task in many mapping organizations
(Petrie and Kennie, 1990; Niemann and Howes, 1991). The
format of a DTM is basically a regular grid of elevation.
Ideally, the values are interpolated from the original set of
points (for example from a topographic survey). In Taiwan,
however, the most complete source of elevation data is from
existing paper maps. The challenge thus becomes one of
efficiently converting a large amount of paper maps into a
digital representation for GIS applications.
At present, DTM data are derived from three alternative
sources, namely ground surveys, photogrammetry and
digitization of cartographic documents (Weibel and Heller,
1991). Ground survey measurements tend to be very accurate
and the accuracy of the resulting DTM is very high. However,
as this technique is relatively time consuming, its use is limited
to small areas and large scale projects. Photogrammetric data
capture is based on the stereoscopic interpretation of overlapped
aerial photographs or satellite imagery. It is normally used in
large engineering projects (e.g., dams, roads) as well as
nationwide data collection projects. Digitization of cartographic
documents may be implemented by manual digitization, semi-
automated line following, e.g. Laserscan" (Peuquet and Boyle,
1984), or by means of automatic raster scanning and
vectorization of a contour plate, e.g. Sysscan* (Peuquet and
Boyle, 1984). The accuracy obtained will depend on the
quality of the cartographic sources as well as the Analog/Digital
(A/D) conversion process. Due to the relatively high costs of
the direct methods of data collection (surveying and
photogrammetry) and the large volume of existing paper maps
available, the indirect method (A/D conversion) is predominant
for large data collection projects. This is particularly true for
national (e.g., USGS) or military agencies (e.g., DMA).
Despite the widespread use of paper maps as a basis for DTMs,
collecting contour data presents its own problems. Essentially
contour lines are a form of terrain visualization and are not
primarily used as a scheme for numerical surface
representation. Excessive numbers of points are sampled along
contours, whereas no data is sampled across contours. Errors
may be introduced in drawing, line generalization, and
* Mention of vendors system should not be construed as an endorsement
of the product.
265
reproduction. A large amount of the original information is lost
in the map making process. Contour generation is a
generalization of spot height information, consequently,
contour data yield DTMs of only limited accuracy. However,
since large area coverage is achieved relatively cost effectively,
digitized cartographic documents provide a compromise method
for creating DTMs for use at medium or small scales.
The contour lines in a topographic relief plate are a graphic
representation of terrain. The focus of the A/D conversion is
the interpretation of this graph. When a graph is stored in a
digital environment, descriptions are required and attached to
the graph for interpretation. For these isolines, the description
is the elevation data it represents. Contour labeling is a process
whereby elevation data are attached to isolines. Traditionally,
vector based techniques dominated the contour labeling process
which relied heavily on high labor intensive manual operation,
costly computer graphics hardware and were time consuming.
As a consequence, these constraints limit its use (Selden,
1991). But the low cost of scanners and large storage mediums
make the raster based approach more popular (Carstensen and
Campbell, 1991). An approach that incorporates the advance
of computer technology, the relatively inexpensive raster-based
scanning machines, the ever increasing computational speed,
and the need to reduce the cost of manpower, to minimize
errors introduced from tedious repetitive manual tagging and to
provide a user friendly environment, is proposed in the
following discussions.
This paper presents an approach that efficiently utilizes a
minimum set of height information from the topographic maps,
to automatically tag the contour lines and consequently creates
DTMs. Height information includes spot heights and contour
indexes. A conceptual model, in the form of a contour tree, is
adopted to express the relationships of the contour lines, in
particular the contour topology. The topology of contour lines
is derived by examining their neighborhood relationship. A set
of rules are employed to orient the contour tree. In the process,
the human-machine interaction will provide relevant feedback
messages to guide the operator. The minimum requirements of
height information for a free tree to be oriented are derived and
provide valuable messages for a human operator to completely
label the contour lines. This approach incorporates the implicit
nature of contours (i.e., their topology), with the explicit
information about the contour plate (i.e., the height
information) in order to fully automate a contour labeling
system.
j
|
|
|
|
i
i
