GIS Postgraduate Researcher
School of Geography & Environmental Science
Monash University, AUSTRALIA
Email: [email protected]
This abstract expands upon the proposal from Ghardirian & Bishop (2002) whereby maps are often termed as models of reality. For many years, cartographers had to face the difficulties of showing three dimensional (X, Y, Z) realities on a flat (X, Y) surface of a map. The concept can be explored in a variety of ways. Firstly, there is a third dimension (height) that varies continuously over space and has to be flattened to a two dimensional map. There is no doubt that three-dimensional textured urban models are growing in popularity in the field of geo-visualization. The term ‘Geo’ refers exclusively to the geographical representation of GIS data in a ‘real world’ coordinate system which differentiates it from a ‘movie world’ arbitrary coordinate system. The increasing development of urban areas demands more now than ever of urban planners, which in turn demonstrate a growing, demand for 3D realistic visualization. While, the research in 3D GIS was progressing, it soon became a 3D spatial project, utilizing photogrammetry as a fundamental data collection technique. In this context photogrammetry is used as both a geometric data gathering tool and also as a remote sensing tool for information extraction. Field surveying prevailed in obtaining control points utilized in aerial triangulation and block adjustment used for processing of the data. Consequently, with resources and knowledge available from three separate but interlinked fields in spatial sciences, a GIS was created while explicitly concentrating on 3D Geo-Visualization.
For many years, the most challenging issue in spatial sciences has been the handling of 3D Geo-Information, which has resulted in development of many software applications. The project aims & objectives comprised the initiation of 3D GIS, where the idea of integrating digital photogrammetry, remote sensing, surveying and GIS has intimidated many within the GIS community. The issues associated with the realization of this concept are discussed and some of the technical steps in applying this to Central Business District (CBD) Adelaide, South Australia are presented. While the research in 3D GIS was progressing, it soon became a 3D spatial project, utilizing photogrammetry as a fundamental data collection technique. In this context photogrammetry is used as both a geometric data gathering tool and also as a remote sensing tool for information extraction. Field surveying prevailed in obtaining control points utilized in aerial triangulation and block adjustment used for the processing of the data. Consequently, with resources and knowledge available from three separate but interlinked fields in spatial sciences, a GIS was created while unequivocally concentrating on 3D Geo-Visualization. The goal of the research was to develop a 3D realistic city model. This was carried out through the creation of footprints, reconstruction of buildings (see figure 1), photo realistic rendering and transformation to perspective fly-through and web based applications with Virtual Reality Modelling Language (VRML). The results of the research contribute to (1) the creation of 3D realistic models (2) updating and managing existing GIS footprints (3) 3D visualization in terms of geographical visualization and (4) storing original measurements on commercially available systems. In addition, the project explores Virtual Reality Modelling Language (VRML). Just producing visualization is not enough; it should be ensured that what is being produced is based on results and facts of appropriate information analysis.
The major conclusions of the project were:
- A 3D Geo-approach for urban object visualization is better than creating simple extruded 2D footprints.
- Stereo photogrammetry can be effectively used to update current 2D footprints.
- It is possible to work within a geographical coordinate system rather than an arbitrary coordinate system.
- Complexity of urban topographic objects presents significant issues for their measurement and rendering
- Proper orientation and triangulation of faces for purpose of visualization of the model is important.
Collecting geographic data is vital to creating and maintaining a GIS, because inaccurate or outdated information will not reflect true, real-world scenarios. The GIS’s of today are built using information derived from various types of geographic data, chiefly vector-based data. The increasing number of new data sources requires new tools that can bring into clear view today’s complex world of 3-D objects, features and spatial interactions, and help build vital infrastructure for the future. The accuracy of photogrammetry is well regarded, but the costs and associated learning curve have forced many GIS managers to select less accurate methods in their production. This research project would be nearly impossible to visualize using conventional software, since many complex urban components are involved. Elements included complex terrain, water, automobiles, hundreds of existing and proposed city buildings, and trees planted along streets and rivers. It was also important to integrate existing GIS, satellite data, and maps describing the existing city. Fortunately, geographic stereo imagery provides the ideal solution for deriving current and accurate 3-D information, and has proven to be a cost-effective tool for updating a GIS. Digital photogrammetry is an essential imaging tool for creating and extracting useful and building 3-D GIS, considering that the detail required for the visualization of buildings and their geographic environment can sometimes be quite intricate.
The research quizzed couple of questions.
1. How detailed does a building have to be for an unfamiliar observer to recognize it?
2. How real (lifelike) computer can generate 3D model, which can further involve extensive labour hours?
Figure 2 shows a comparison of hand held photograph with computer generated model. Trees are explicitly planted as textured models and are available in different model formats i.e. dxf (3d studio format). Texture mapping in one way to achieve realism than this can be argued to what should be included in the whole virtual world. Everything as it exists in the world should be mapped and textured. A process which is very time consuming and labour intensive. In the movie world the making of such a world in front of the blue screen cost millions of dollars. These films (i.e. Sherek, Lords of the rings, etc.) are fictional and mostly in arbitrary coordinates and projected in front of the screen. Output products created by 3-D geographic imaging techniques include orthorectified imagery, DTMs, DSMs, 3-D features, and spatial and non-spatial attribute information associated with a geographic feature. Using these primary sources of geographic information, additional GIS data can be collected, updated and edited. DSMs created from high-resolution imagery are useful for Identifying and categorizing urban and rural land use and land cover. 3-D topographic information such as slope, vegetation type, soil characteristics, underlying geological information and infrastructure information can be collected as 3-D vectors.
Urban visualization projects often have multiple goals, and it is important to clarify these as early in the process as possible, because it is technically and practically impossible to meet some of these goals simultaneously. One important distinction is between working documents and representations of a final product. Most of visualization done in the design fields is to present a final design to a client. However the research in this project was to develop better working tools, methods and understanding.
A classic strategy used by most professions explores the development of working methods using very efficient, specialized terms and symbology. For instance, much terrain grading work is done using contour maps. The difficulty with this strategy in landscape design is that many professions are involved, and they do not share terminology or symbology. The strategy parameters influence the success and accuracy of the matching process. Furthermore, this visualization technique is an excellent tool for understanding urban development in the past and running prediction models for the future. In this research, the strategy of the project was to develop “realistic” (see figure 2 and 3) representations of each major type of terrain and building object. The depictions were mostly self-explanatory and suggestive; a building looked like a building, and not a polygon; trees were represented using appropriate models, etc. No legend was required, and any thoughtful layperson could give useful feedback about a design.
While there have been advances in the creation of realistic representations of scenes and the display technology itself, the reliability of the results have not been explicitly tested. The novelty of this project lies in the combination of environmental measures, high quality computer graphics, and measures of human perception which serve to validate the fidelity of simulated critical environments.
The top-down approach allows more details about the roof to be collected but requires longer and complex processing of data [Brunn et al, 1996] and [Hendr, 1997]. The true effect in top-down approach is processed with two overlapping images (a stereopair), or photographs of a common area which are captured from two different vantage points which are rendered and viewed simultaneously. Photogrammetry requires a minimum amount of ground control, and number of GCPs will vary from project to project. Six GCPs were collected over a stereo pair for quality control. Five GCPs (Table A) were used in conjunction with block bundle adjustment to establish a geometric relationship among the images, the sensor model and the ground, so accurate 3-D data can be collected from imagery.
Tie points were used to create geometric harmony among the images so they’re positioned correctly relative to one another. Block bundle adjustment (triangulation) was essential in determining the information required to create orthophotos, DTMs, digital stereo models (DSMs) and 3-D features. The orthorectification process will remove geometric errors inherent within photography and imagery. This is only true at ground level, with the tops of vertical features like buildings still geometrically wrong. Although it can be fixed by “True Ortho”, however the project used ordinary orthos. Measurements and geographic information collected from an orthorectified image represent the corresponding dimensions as if they were taken on Earth’s surface. 3-D data and information can be collected from Digital Stereo Models (DSMs). Where DSMs will allow GIS users to interpret, collect and visualize 3-D geographic information from imagery and is used as the primary data source for collecting 3-D GIS. The rendering process in the 3D visualization is usually solved via the use of 3D triangles and the add-on of colour, tone, typical imagery, or actual imagery of the object. The true effect in top-down approach is achieved with two overlapping images (a stereopair), or photographs of a common area are captured from two different vantage points which are rendered and viewed simultaneously. In such a case an X parallax errors is reported and if orientation is not done properly Y parallax error is found. Fixing such an error is very important in manual digitizing. Implementation to fix such an error is reported illustratively in figure 3. Notice the cursor in red on both figures. In figure 3A ‘left image’ the red cursor doesn’t align or rest at the same place as cursor in the right image. This image has to be corrected for x-shift as it reports an error in z-value or position. Notice the shift in x-direction with blue arrows as illustrated in Graph A-C. The graph in figure A shows a huge shift in x-direction while the graph in figure B has been reasonably corrected resulting and perfectly positioning Graph C. This therefore corrects the image in figure 3B. However some errors are easily pointed as it should superimpose on the same point as on the right image to get perfect stereo effect and precise position.
Note: Graph C shows perfect example in x-parallel error correction, where a cursor lies exactly at the same point in both images. This is an ideal condition to follow with top-down approach and digitizing rooftops.
The research in photogrammetric software development has come a long way. Fischer et al,  and Forstner  reported using application where the user has to define the building model and find the building elements in one image by mouse clicking. The algorithm finds the corresponding features in other images; and matches them to build the 3D wire frame of the building. This approach supports the extraction of more complex building; however it requires the user to spend more time interacting with the system. Current methods that aim at automating feature extraction require manual work for grouping and cleaning the data set [Hendr 1997], or for matching and fitting predefined shapes [Brunn et al, 1996], [GrünA 1996]. The more complex the topography and the higher the required resolution of its model, the more superior is human interpretation of the stereo model.
This application delivers 3D perspective walk or drive-through. Many movie-like simulations of travel through a 3D virtual world are pre-computed with the path being fixed. In this application there are a very large number of from – to solutions. Location is solved through use of the internet and 2D GIS maps to define source (GPS in a vehicle is an alternative here) and destination. Route optimization occurs once observer parameters are defined. High performance computer(s) provide a solution for the walk or drive along the optimized route and this is delivered to the observer.
Results indicate that a) architecturally simplistic buildings, which present flat facets to an observer, are much easier to measure and render than architecturally complex buildings; b) whilst it is important to render in some detail visual landmarks, significant effort can be saved by using simple tone or typical textural rendering of less significant features; c) it is practically difficult to accurately render all sides of 3D objects; d) greater realism is achieved when familiar objects such as trees and vehicles are added to the 3D model; and e) maintenance of geographic coordinates is difficult in existing geo-visualization software if quality visual enhancement of 3D objects is desired.
VRML browsers are generally known as means of visualization of 3D graphics on the Web (cosmo player), allowing real-time navigation. In this research project, reconstructed buildings and related features are coded in VRML file (see Figure 5) that opens in normal Internet browser. If the user wants to click on a building, a sensor has to be attached to this building in the VRML document. If the user wants to have the animation of a walk along a street, the route and the speed of walking have to be specified in a special VRML node. Linking 3D VRML environments to a relational database is a challenge. The project also demonstrated that interactive geo-query tools are used for querying 3-D visualizations, enabled by seamless links to GIS-based attributes, providing a framework for a suite of 3-D landscape design capabilities.
An IBM ViaVoice embedded system was tested to monitor its interaction with windows platform and it was noted further improvements are needed. Windows Developer Kit can be used to write algorithms so that the application (1) will integrate with GPS system, (2) the system is integrated with a database in order to map real time coordinates and (3) the user can interact with touch screen or voice command recognition. Software share several key capabilities that go beyond the requirements for vastly improved performance. Such capabilities include: (1) 3-D object rendering of symbols such as trees, buildings and other manmade features. (2) Texture mapping to support realistic rendering of polygonal features such as roads, meadows, open water and background sky and (3) Wireframe upgrading from triangle snapping geometry to user defined octagons, rectangles etc.
In this project application, car wireframes and trees (see figure 5) are shaped and added to affix more realism to the environment. The research project demonstrated that 3D Geo-Visualization of urban environment is not a simple process, but is in fact a whole new dimension of research and visualization. It is argued that these psycho-perceptual issues require careful consideration as they can significantly influence the cost and effort of generating data.
- [Ghardian & Bishop] Payam Ghardirian and Ian D. Bishop, AURISA 2002. “Composition of Augmented Reality and GIS to Visualize Environmental Change”.
- [Fischer et al, 1999] Fischer A., Kolbe T.H., and Lang F. (1999). On the Use of Geometric and Semantic Models for Component-Based Building Reconstruction Proceeding. Semantic Modelling for the Acquisition of Topographic Information from Images and Maps, Smati ’99 Workshop, Institute of Photogrammetry, Bonn University, pages: 101-119.
- [Forstner 1999] Forntner W. (1999). 3D City Models. Automatic and Semiautomatic Acquisition Methods. Photogrammetric Week, Stuttgart.
- [Brunn et al, 1996] Brunn, A., E. Gülch, F. Lang and W. Förstner: A Multi-Layer Strategy for 3D Building Acquisition, in: Proceedings of the IAPR TC-7 Workshop, Graz, Austria, September 2-3, pp. 11-37, 1996
- [GrünA 1996] Grün, A.: Generierung und Visualisierungvon 3-D Stadtmodellen, in: Proceedings of the IAPR TC-7 Workshop, Graz, Austria, September 2-3, pp. 183-196, 1996
- [Hendr 1997] Hendrickx M., J. Vandekerckhove, D. Frere, T. Moons and L. V. Gool: 3D Reconstruction of House Roofs from Multiple Aerial Images of Urban Areas, in: IAPRS, Vol. 32 Part 3-4W2, Stuttgart, September 17-19, pp. 88-95, 1997