PHOTOGRAMMETRY
The Third Dimension
Exploring object-oriented modeling in stereo photogrammetry
By Simon Fletcher and Susan Goodman-York

In recent years, the technology for surveying, remote sensing, digital mapping and GIS has grown dramatically in both depth and diversity. However as a result, the methods and workflows for the capture, management, analysis, display and production of geospatial data have become increasingly fragmented and disjointed. Only recently has object-oriented geospatial technology overcome the difficulties in effective integration of raster and vector data. Even so, this integration has, until now, been largely limited to 2-D datasets.
    New innovations have dispelled these limitations. Real-time, 3-D update of an object-oriented geospatial database using photogrammetric software is now a practical possibility.

The Geometry of the Object-Oriented Database and its Management System
In traditional GIS, spatial data are held in a vendor's proprietary format while the non-spatial attributes may be held in a relational database. This complicates data management and queries. With an object-oriented approach, the spatial data can be held directly in the same database as the non-spatial data. This significantly enhances the system's ability to analyze, process, manipulate, display and store spatial information.
    Traditionally, geospatial databases represented real-world features such as roads, rivers and lakes as linework with attributes. This makes it difficult to consider a river as having different geometry at different scales, or to allow features in the database that have no geometry, such as bus schedules and other relationships.
    The object-oriented data model, on the other hand, is capable of representing and managing more true-to-life structure and relationships among real-world entities. In an object-oriented environment an object is no longer an attribute with a unique geometric description, or, vice versa, geometry with an attached description. An object in this environment encapsulates within itself values, structure and behavior.
    Objects within the database are essentially intelligent and can adjust themselves according to changes in the geographic environment surrounding them. Methods for self-validation of data can prevent topological inconsistencies. Objects can have multiple representations for scale-dependent and cartographic display. Furthermore, the object and dataset versioning support multi-user update with full audit trail capability.
    Object-oriented databases and their associated database management systems (OODBMS) enable objects to be treated both individually and collectively. Most commercial OODBMSs, however, have not addressed certain crucial issues of GIS, such as efficient spatial indexing, geometric-topological structuring, and multi-user long transactions (see sidebar). Combining all these capabilities into a coherent OO-GIS results in a fundamentally new generation of database management system.
    Object properties can be defined in many ways to include the object's attributes, relationships to other objects, and geometries (coordinates). The system departs radically from the traditional method of holding 2-D geometry (X,Y or latitude/longitude) and object height information separately; instead, all this information, including contour values, sounding depths and spot heights, can be recorded as object attributes.
    Additionally, raster data containing the height information for a whole area - for example, a Digital Terrain Model (DTM) such as U.S. Geological Survey's (USGS) Digital Elevation Model (DEM) or U.S. National Imagery and Mapping Agency's (NIMA) Digital Terrain Elevation Data (DTED) - can be held simultaneously with vector objects, enabling the user to access the most appropriate form of height information for the task in hand. Recent advances in object-oriented technology have extended the vector-object geometry capability, enabling an object to hold per-point height information (Z-values). This opens up new avenues for building and maintaining true 3-D datasets.

The Solutions
Earth scientists and Earth observation specialists often take advantage of photogrammetry as a data source, whether this is for the purposes of updating databases or for initial data capture. Legitimately, they also want to guarantee the quality and integrity of the data captured, and need a proven data management system.
    The need to extend the level of data integration to enable real-time 3-D database update has been addressed by Laser-Scan and LH Systems in the integration of Laser-Scan's Automated Map Production System (LAMPS2) and LH Systems' Helava SOCET SET¨ photogrammetric software suite.
    LAMPS2, a digital map production tool driven by a fully object-oriented database management system, and SOCET SET, running on a modern UNIX or Windows NT platform, is a state-of-the-art stereo digital photogrammetric workstation. In this integrated environment, it is possible to view stereo imagery, selectively overwriting 3-D information from the existing LAMPS2 database. The user can then update, edit and validate the information from the stereo photogrammetry platform, feeding this back to the object-oriented database in LAMPS2, all in real-time. Since the system is built upon an OODBMS, all of the benefits of object-orientation are available. For example, topology validation methods already built on the objects in the database ensure correct topological structuring in real-time while the operator is digitizing features from the stereo workstation.

Downsizing the Warehouse: the Economies of OO-GIS
Today's photogrammetrist faces the new problem of being confronted with information overload. Higher resolution satellite imagery, along with newer means of data collection (SAR, IFSAR, LIDAR) provide terabytes of information to build the photogrammetric model or map. The efficient automation of this process, however, requires a system with the ability to handle large and complex vector datasets.
    This volume and complexity of data create a data warehousing nightmare. For instance, in its attempt to better understand how coastal regions are changing due to erosion and other factors, the U.S. National Oceanic and Atmospheric Agency (NOAA) is now using an Airborne Topographic Mapper with LIDAR technology - capable of determining elevation changes down to 10 cm - to map and model coastlines. Placing all of this information into a DTM and then comparing it with future coastal changes is a daunting task, but an OO-GIS would allow individual features or objects to be compared and updated much more rapidly.

Object-Oriented Technology and Stereo Photogrammetry: A Conclusion
Object-oriented GIS technology enables the user to model a world that is dynamic and changing - in other words, the real world. The technology copes with and adapts to changes in context and scope of the data. It supports a more intelligent modeling of the complex entities and relationships in geospatial data which evolve as our understanding deepens and decision makers' needs change.
    In photogrammetry, imagery and vector data are close bedfellows. In the object-oriented world-view, data representation and data manipulation are part of a single continuum. LH Systems and Laser-Scan have shown beyond doubt that both disciplines have a great deal to impart to one another.

About the Authors:
Simon Fletcher and Susan Goodman-York are members of Laser-Scan's international marketing team.

Back