Images for Intelligence
TEC and the new alliances between photogrammetry and mapping
By S.J. Fletcher

The relationship between photogrammetry and map production is evolving rapidly. Only recently has photogrammetric feature extraction been seen as a stand-alone activity; it might eventually feed into mapping and GIS, but it is far from an organic part of these processes. A lot of off-line work had to be done to turn what was caught on film into a geographic resource.
    This traditional view is now being challenged. A new technology is available which brings together the previously separate disciplines of photogrammetry, data modeling, and cartography, resulting in a unified flowline for the production of maps, charts, and the geospatial data. The key concepts in this technology are synergy and dynamism; it is a totally interwoven process. Images can update the database. The database can overlay features on existing imagery. You can view the results of the modeling in 3D or as a cartographic representation. Uniquely, the whole process occurs within the same data model, on the same screen, via the same user interface.
    This is a considerable improvement on what went before. In the past, the photogrammetry operator had to digitize features, then export the linework in a standard commercial format such as DGN or DXF, then import those files into a GIS to generate topology. This was an unwieldy and time-consuming process, but all too often it was only the beginning of the headache: the topology generation process could alter the original geometry, and to validate these changes the linework had to be passed back to another specialist.
    This rather self-frustrating cycle of digitization and repair is totally avoided in the new technology, because, as we shall see, the photogrammetric system and the geospatial database driving the GIS are essentially 'wired' together; topology generation and validation occur at the time of digitizing, enabling the imagery specialist to identify and fix any problems directly and immediately.
    The technology has attracted interest from the United States Army's Topographic Engineering Center (TEC), which is now using it in ongoing mapping research with specific reference to US Department of Defense requirements. The installation currently comprises two seats, but provision has been made for up to thirty licenses for simultaneous user access.
    The software which promises to so change the complexion of image-based map production, consists of a combination of a stereo photogrammetric workstation (LH Systems' SOCET SET) and a map production application (LAMPS2, produced by Laser-Scan, Inc., in Sterling, Virginia, and driven by that company's Gothic geospatial database).

Photogrammetry: Its Place In Map Production
TEC supplies topographic and other geospatial information to the US Army. Providers of maps and geospatial data now face a complex set of opportunities and challenges. Users want information in a diversity of hard- and soft-copy forms; affordable, validated, and up-to-date. Imagery from satellites and aerial photography, as an alternative to re-survey, is a cost-effective source for much of this data.
    Stereo imagery works best in the process; it not only provides information for 3D datasets, but also enables better image interpretation for the production of 2D data, maps, and charts. However, simply obtaining the imagery is not the key to effectively completing the process. The imagery must also be exploited to enable the organization to meet multi-product requirements economically and efficiently.
    This entails extracting features and giving them topological structure, rules, error-checking methods, and other geospatially intelligent refinements so that they become geospatial objects with their own behaviors and reactions, rather than simply points, lines, and areas.

TEC: Objectives And Technical Requirements
TEC's technical requirements were defined in three basic categories: topological feature extraction interface, edit environment, and data structure support. These cover the processes which make up the flowline. The following sections explain TEC's requirements and briefly show how the LAMPS2/SOCET SET features have satisfied them.

Interface and feature extraction
The interface between the two software products was to provide a topological, object-based feature extraction capability. The extracted features were to be displayed over the stereo imagery in the SOCET SET real-time display and stored in the Gothic active object database underlying LAMPS2.
    The interface was to enable the system to support 3D coordinates (x, y, z) on feature geometry (point, line, area), and 2D coordinates (x, y) on topological primitives (node, link, face). It involved the creation of new base classes within the Gothic database (simple point 3D, simple line 3D, and simple area 3D), to accommodate the new dimensional data. The user's real world feature classes can 'inherit' behaviors from these base classes, and these can then be used to hold the vector data generated by SOCET SET.
    For example, the object classes expressway and secondary road can both inherit from a class called road, which inherits in turn from the simple line 3D base class. This 3D base class imparts the behaviors and characteristics appropriate to an object with 3D coordinates. The structure of the classes is also carefully arranged so that the topological information enables expressway and secondary road to behave as part of a network for analysis.
    Representing point, line, and area features with 3D geometry demonstrably reduces ambiguity. For example, two roads that intersect can be represented with an elevation value at every vertex, and their heights are distinguished if one passes over the other.

Edit environment
The main editing development in the TEC project was the creation of a 3D edit frame to operate on objects with 3D geometry stored in the Gothic database. This frame was to be capable of interactively editing 3D point, line, and polygon feature data using both spatial and attribute edits. The edited features would then be graphically superimposed over stereo imagery.
    These features are currently necessitating some extensions to the functionality of the Gothic database, to ensure that the extensive data manipulation possibilities offered by the database-editing, digitization, customizable display representation, and other refinements-are also able to work with 3D geometry.

Data structure support
Some of the objects which TEC's dataset might feasibly be expected to contain have special 3D topological requirements, and the system is currently being configured to manipulate these. For instance, some objects may require the storage of multiple attributes of various types against a single vertex within an object's geometry. A common example is that of a cliff, where both absolute z values, as well as values for height above ground or sea level, must be recorded. As well as storing z values, this capability can also be used to store other types of per-point attributes.
    Work is also being carried out to enable consecutive vertices in a 3D geometry to have identical (x, y) coordinates, which is normally forbidden in conventional GIS. The point of this is that, in large scale mapping, a building, for example, might have different roof sections at different heights. The normal (x, y) coordinate restrictions could be problematic in representing this kind of topology.

Gothic In SOCET SET: The Advantages Of Active Objects
Gothic's active object approach to mapping will bring a number of enhancements to TEC's use of SOCET SET.

Data modeling and update
Gothic in SOCET SET enables the user to interactively update the data schema at any time during editing. It is not necessary for the user to define every object class and its attributes before beginning feature extraction.
    Gothic also provides validation methods that prevent inconsistencies and errors such as self-intersecting areas. Yet the flexibility of the data model means that it is perfectly possible for a single object class to be defined as point, line, and area, depending on how and where that object appears. Thus, a bridge, for example, can be a line in one location and a point in another location. There is no need to define three separate classes for area, line, and point, which is the only way a non-active object data model could cope with this polyvalence.

Topology and data cleaning
Topology is built automatically in Gothic as SOCET SET is being used. It is not necessary, for example, for the user to manually place nodes, as would be the case in less intelligent data models. As an extension of its data validation capabilities, Gothic also checks for overshoots in the data and can detect gaps and slivers between objects.

Editing
The straightforward editing tools in Gothic will enable TEC to perform operations more quickly and with fewer mouse clicks. For example, when digitizing a line, the cursor automatically snaps to the stereo screen. To move a point, the user simply clicks on the vertex to be moved and drags it into its new position. Over hundreds of transactions, this reduction in the need for mouse clicks will be very clearly felt.

Representation
As always, representation is a crucial issue in mapping. Gothic in SOCET SET provides the ability to produce chart-like representations of data. The user can capture the data and view it simultaneously as it will appear on the chart via the LAMPS2 display. If the user wishes to change the way a particular object is represented, these changes can be made globally to all the objects in the appropriate classes. There is no need for the user to change representation on one feature at a time. Thus, TEC can make symbols representing their allies' command posts a different shape to their own, or 'switch off' narrow roads when planning routes for heavy armor.

Hard-copy map production
The SOCET SET/LAMPS2 combination is not merely a pre-production data processing device; it can also produce cartographic quality hard-copy maps and charts from the digital data which it has compiled. All map environments demand that the data which underlies them be managed as uniformly as possible; the LAMPS2/SOCET SET combination will enable TEC to set up a two-way update and compilation process between imagery and database and distribute the results of this process as portable map sheets.

TEC's Use Of SOCET SET/LAMPS2
It is difficult at this stage to give precise details of TEC's use of the SOCET SET/LAMPS2 combination; the project is essentially one of research and development and many different avenues are being explored.
    Confidence in the two-way, dynamic imagery-database approach to mapping is, however, high in the organization; as TEC physical scientists Dan Edwards and Justin Simpson commented in their recent paper, given at the Laser-Scan User Group, "This collaborative effort has permitted government and industry to work together to develop an essentially real-time topologic 2.5D mapping and database capability. The development of a framework for the integration of these two separate technologies holds great promise for greatly improving feature mapping efficiency and the quality of spatial feature data."
    In addition, the Gothic database technology behind the SOCET SET/LAMPS2 combination is already in use in other military organizations throughout the world, including the Spanish and British armies. The active object approach has already shown itself to be a proven and robust technology in these applications.
    The future of mapping belongs to those technologies which can seamlessly combine all the underlying data-survey, re-digitization, imagery-and manipulate them accurately as one. Creating a dynamic environment where photogrammetry benefits from active object intelligence, and the active object database is populated using imagery, is a vital step in developing this environment. TEC's work will help to ensure that the interface between these two technological elements will become sufficiently honed to produce a true end-to-end, image-based map production process, specific to defense needs, but at the same time adaptable enough to address the entire spectrum of cartographic challenges.

Back