|
Peter Becker describes a methodology to structure
photogrammetric data suitable for both 3D CAD and GIS
Overview
Topologically-structured GIS data have a considerable
advantage over conventional CAD or line mapping data in that they
deal with inter-related areas (houses, land parcels, courtyards,
etc.) that possess common boundaries. Not least, such data can be
interrogated and manipulated for the purpose of:
Establishing answers to such questions as, "What buildings
partially extend beyond approved plot boundaries" or, "Which
buildings do not conform to zoning categories"
Deriving area statistics, e.g.,
What proportion of land has sealed surfaces, etc.
Displaying/plotting map data with area based solid fills.
Well-structured map data should avoid redundancy. In this context,
the same geometric element that defines, say, the edge of a pavement
could also define the edge of a garden where this is applicable.
Not adhering to this principle undermines data integrity, which
in-turn makes data maintenance uncontrollable.
Drawbacks
Despite these advantages, conventional GIS
data are difficult to maintain and a topological structure generally
imposes numerous limitations on the data set, e.g. no 3D, no splines,
no curves. As a result, photogrammetric 3D data is often converted
to simplified 2D data sets. Such GIS data sets are inherently unsuitable
for 3D CAD geoengineering applications such as road design or detailed
planning and where curves, splines and 3D are essential to maintain
the required level of accuracy and data manageability.
Most large scale mapping is created and maintained
photogrammetrically in 3D. Advanced map updating is performed by
digital superimposition of the old data over the new situation as
displayed in a stereo model - an impracticable procedure with 2D
data. This, alone, is a strong point for maintaining data in a true
3D form.
The importance of 3D data is steadily increasing
in line with applications that make use of those data. The need
for accurate 3D city models is a particular requirement of cellular
network operators, and 3D modeling is now becoming an affordable
tool in general site and town planning schemes.
Reconciling differences
Digital vector mapping systems that have emerged
to date have simply replaced their analogue predecessors (i.e. mechanical
stereoplotters). Such systems make use of CAD to assist the drawing
process, but do not exploit the power of topology available from
GIS.
In addressing the problem, MAPSgeosystems
realised that a conventional approach to photogrammetric data capture
would not overcome the seemingly conflicting requirements for mapping
data. i.e. data possessing the following virtues:
be 3D
be fully topologically structured
allow the use of curves and splines
allow database linkages for storage of attribute information
be storable as standard file formats or within an RDBMS
have reduced complexity
be easy to maintain
be multi-use and usable in multiple environments
be cost efficient
Significantly, the company had, from its creation in 1986, fielded
a strong Systems Engineering department to exploit CAD/GIS technology.
The fruits of its labours were evident in such packages as MicroMAPS
on MicroStation and AutoMAPS on the Auto-CAD platform, as well as
tools for on-line topology checking, snapping, dynamic segmentation,
image processing, etc. Experience from many hundreds of mapping
projects worldwide has also been applied to satisfying the criteria
listed above.
Primary Data structure
The concept of Primary Data - one whereby
data are captured and maintained to the lowest common denominator,
free of redundancy - has always been maintained at MAPS. In practice,
the presentation data are not stored, but derived from attributes.
For instance, a manhole is stored with all its available attributes.
Some attributes, e.g. size and rotation, may be used to represent
the manhole at large scale, but not at other scales.
It is often necessary to employ different
symbology during different stages of data acquisition. For example,
when using color images in systems with vector overlays , it is
difficult to identify green symbols over vegetation and better visual
identification can be achieved using alternative colors.
The resymbolisation of Primary Data is used
to define colors, widths, symbols and styles for different production
steps or client-required products. In instances where a product
cannot maintain all the required geometry and attributes, it is
simply generated, used for the required application, and discarded.
Only the Primary Data are maintained.
Such attribute-based symbolisation is a key
component in most GIS systems but is often not maintainable by CAD
systems.
Fugro MAPS Topological 3D Primary Data Structure
To utilise 3D, curves, splines and topology
during data capture and analysis, a revised data structuring method
was required.
This was achieved by:
Enforcing the Primary Data concept
Using topological GIS techniques during data capture
Enabling extended feature coding
Using an intermediate 2D polygon by-product
Development of automated routines.
The geometry of a Topological 3D (T3D) Primary Data Structure is
stored as 3D CAD elements with multiple feature codes. The latter
are categorised as:
Topologically significant:
Boundaries: Features that separate two areas
Centroids: Identifier for an area defined by enclosing boundaries
Non-topologically significant:
Points: e.g. poles, spot heights
Lines: e.g. power lines
Edges: Features derived from areas (e.g. edge of garden)
Areas: Polygons captured as such (e.g. tree boundaries) or derived
from the topology
Text, including text for cartographic placement
A single recognisable object can be defined as more than one feature
type. For example, a wall can exist as a Boundary Feature (wall
between two courtyards) or as Line Feature (e.g. wall within a courtyard).
Feature codes are also classified as being
Primary or Presentation. The former define the geometry of real
objects, while the latter are generated from Primary features for
presentation purposes. For example, a spot height point is a Primary
Feature, while the associated text label is a generated Presentation
Feature.
Geometry is maintained as 3D CAD elements
using the above structure. The sub-sets of topological features
are maintained in a topologically clean form so that when dropped
to 2D and simplified to linear elements, the GIS can build topology
with no errors.
Revisions are always applied to the Primary
Data, from which all other data are derived.
Data capture methodology
The data are captured photogrammetrically
as 3D features consiting of boundaries, centroids, lines, points
and areas using the available element types. Instead of capturing
individual boundary objects with separate featured codes, many such
features can be captured as general boundaries, thereby speeding
the task. Line or boundary features (e.g. fences) that require specific
feature coding, as well as spot heights, breaklines and point features
are also captured. Feature coded centroids for all bounded areas
are added in 3D at the surface elevation, or in 2D from approximate
ortho images.
To assist such exercises, on-line topology
checking and segmenting routines developed by MAPS are used to dynamically
control logical snapping, ensure no dangles exist in this 3D data
set, and perform 3D segmentation.
Following initial data capture, the topological
subset is dropped to 2D and processed through clean-up routines
in a GIS engine (e.g. Geographics or ArcInfo). Such routines return
flags warning of integrity errors such as undetected dangles, areas
lacking centroids or unused boundaries. The error flags are superimposed
on the Primary Data and manually corrected. The processing is repeated
until the Primary Data is topologically clean.
Polygons are created from the clean data
by a GIS engine and attributes and database linkages of the centroids
(e.g. elevation) are transferred to them. These polygons form an
area based model which has many uses. It can also be used in a CAD
system to create solid fills or be imported into GIS applications
for quick analysis or 3D visualisation.
The polygons are used in a special program
to re-feature code the Primary Data. In this step, up to three feature
codes are added to boundary features for polygons to the right and
left of the feature. Thanks to this process (enabled by MicroMAPS),
a single linear element can represent a fence, the edge of a garden
and edge of pavement.
Priorities are given to feature codes in
a way that allows a feature to be symbolised according to its priority.
In the previous example, priorities would be set so that fence has
the highest priority, followed by edge of garden, followed by edge
of pavement. Turning off fence results in the line being symbolised
as edge of a garden.
Logic consistency checks are also run on
the data. These automatically assign codes to features missed during
data capture and check for features that are not present in reality.
For example:
Features with codes: Sea & Sandy Area
& General Boundary should be changed to Coast Line & Sea
& Sandy Area.
Features with codes: Sandy Area & Courtyard
& General Boundary should be checked as the General Boundary
is likely to be a wall or fence.
Raw data capture
Multiple feature codes assigned to boundaries
Products
The result of this capture methodology are
TAD Primary Data defined by CAD element and free of redundancy.
These data are maintained and used to automatically generate a number
of derived products.
One such product, generated from the topologic
subset in different formats, is the 2D Polygon file. This is distributed
with the Primary Data and can be used for a number of applications
such as visualisation and statistical analysis. Other derived products
include: Digital Terrain Models, simple 3D city models. 3D virtual
worlds, plus GIS-specific and Engineering-specific data formats.
Simple 3D visualisation from extruded lines & polygons
T3D Primary data overlaid on 2D polygons
Conclusion
MAPS has used the above procedures in a number
of projects, including the mapping of Sharjah city in the United
Arab Emirates.
Its advantages are numerous (see checklist)
and satisfy the criteria outlined earlier. As such, it offers a
viable solution to the problem of creating and maintaining large
scale mapping for use in geo-engineering applications.
Indeed, organisations have much to gain by
specifying that this principle be employed when issuing tenders.
T3D Primary Data checklist
Being topologically clean, such data can be
used directly in GIS for complex spatial analysis Is easily maintainable
from photogrammetric or ground surveys
The data structure need not be clean to be a valid file. It can
therefore be edited by those without knowledge of topological data
structuring and cleaned up at a later stage by those proficient
in such work
No topology is stored in the data, hence complex file formats are
not required and the data can be held as CAD files (e.g. DGN, DWG)
or within relational databases (e.g. SDE, iSpatial). Attributes
can stored with, or linked to, the geometry
Nearly all GI systems can read the data without translation
Derived data sets in multiple formats can be automatically generated
for spatial analysis or visualisation
A standard derived product is the Area Polygons, used for solid
fill plotting and/or areas-based visualisation.
The topological structure enables automated Quality Control and
consistency checks on the data.
The polyvalence of the data makes them applicable to a much wider
range of uses, thus enhancing their value.
Last but not least, the high degree of automation makes data capture
and maintenance highly cost-efficient.
Biography
Peter Becker
Director of systems engineering at MAPSgeosystems
which has offices in Europe, the Middle East and Africa and production
facilities in Germany, United Arab Emirates, Lebanon and Portugal.
He can be contacted by email at: pbecker@maps-geosystems.com
|
Profile
