. 1998 , and to produce a DTM of the complete
country in a grid format of 1 point per 16 m
2
ŽMinistry of Transport, Public Works and Water .
Management, 1997 . Obviously, these types of appli- cations are less demanding than that of producing
information for road planning and design. Although all applications require filtering of both blunders and
Ž undesired objects Fritsch and Kilian, 1994; Pfeifer
. et al., 1998 , the extraction of information from laser
measurements for the purpose of road planning and design requires more sophisticated procedures to re-
Ž cover and identify the objects on the terrain Weidner,
. 1996; Haala and Brenner, 1997 .
In the following sections, we will first present briefly the current practice of DSM production for
road planning and design at Rijkswaterstaat. In addi- tion, we will determine which type of information,
both semantic and geometric, can be extracted from high-density laser measurements. The planimetric and
altimetric accuracies of the laser data are also as- sessed by using reference measurements. The article
finishes with some conclusions and recommenda- tions.
2. Current practice of DSM production for road planning and design at Rijkswaterstaat
At the Survey Department of Rijkswaterstaat, the terrain relief is extracted by measuring pertinent
morphologic features. These features consist of both single points, and strings of points on flat surfaces
and on places where the terrain slope or curvature
Ž .
changes abruptly category 1 in Table 1 . Attached to the coordinates of these features is also semantic
information that classifies the feature, e.g., as a single point or a breakline. This semantic informa-
tion is however not further used, except to compute a
Ž .
triangulated irregular network TIN . The recovery of the objects on the terrain relates to the determina-
tion of the co-ordinates of specific points as well as to their semantic information explicitly required by
Ž .
the user category 2 in Table 1 . The DSM produced at the Survey Department has
thus two different types of semantic information. One type is relevant solely to recover the terrain
relief with sufficient accuracy, while the other is explicitly requested by the user. This means that, if
laser scanning allows the extraction of terrain relief with the required quality and efficiency by measur-
ing a dense set of points, only the second type of semantic information is required.
The DSM is presently generated by photogram- metric means using colour photography at a scale of
1:4000. The user requirements in terms of accuracy are 25 cm in planimetry and altimetry for both the
planning and design phases. In case of hard struc- tures such as roads, and for the purpose of road
design, the altimetric accuracy is required to be 7.5 cm. Although photogrammetric DSMs have been
used to plan and design roads for several years, it is not yet known if it has the required accuracy. This is
because the DSM accuracy is a function of not only the measuring precision, which is known, but also of
the sampling and modelling strategies. The sampling strategy has a strong human component. Although
Ž .
the features to be measured are listed as in Table 1 , and there are guidelines to help the operator in
Table 1 Information content of a DSM produced at the Survey Department of Rijkswaterstaat
DSM features Category 1: terrain relief
Category 2: objects on terrain Ž
. Slopes: upper side and lower side described by breaklines
Separation line between two paved surfaces Ž
. Flat surfaces described by strings of points
Separation line between paved and unpaved surfaces Ž
. Highly curved surfaces described by strings of points
Lines painted on roads Individual points
Protection walls: dam, quay Water boundaries
Ditches Roof gutter and ridge
selecting such features and the number of measure- ments per feature, human interpretation always plays
a role. The modelling strategy relates to the function Ž
used for interpolation and to the data format e.g., .
grid or TIN . At present, a study is being carried out to assess the accuracy of photogrammetric DSMs by
using tachymetric measurements as reference.
3. Study area and data