Visualisering van veranderingen in het landschap : een computer-ondersteund ontwerpinstrument voor de (landschaps)architect

Important parts of a landscape architect's work are designing new developments in the landscape, predicting the visual impact of future actions, and advising on how to fit new elements into the environment. To perform these tasks, he must know how certain design decisions will change the landscape. The visualization tools available today, however, do not permit the efficient manipulation and realistic display of these changes.In a joint research project, the Department of Computer Science of the Netherlands University of Agriculture and the Institute of Forestry and Landscaping "De Dorschkamp" examined the usefulness of the computer for visualizing future developments in landscapes.The research project was done in three stages:- Stage One, in which the most appropriate visualization technique was selected, the hardware and software requirements were specified, and the necessary hardware and useful software were obtained;- Stage Two, in which the necessary software was detailed and the missing parts were written and integrated with the available software into a working system.- Stage Three, in which the system was tested on a few case studies and the software was refined.Capabilities of computer aided-design (CAD) systemsDuring the first part of the research project, the CAD systems that were in use in 1982 were compared to determine how well they could visualize changes in the landscape. Most CAD systems focus on the technical aspects of the design process; consequently, the drawings produced with these systems usually lack aesthetic qualities. The systems are not capable of an efficient input and clear display of natural elements such as trees and shrubs. An enormous amount of data is needed to generate accurate perspective views of the surrounding landscape, making them unfeasible for visualization only. Besides, on most graphics stations, It is not possible to generate a perspective view of something as complex as a real landscape within a reasonable amount of time, unless the images are strongly simplified.More realistic images can be obtained by mounting computer-generated projections of a plan onto photographs of the site, which also eliminates the necessity of inputting landscape data. But, composition is usually done by hand, and adjusting the colours of the plan to the photograph and removing parts that are hidden by the surrounding landscape is cumbersome. Photo montage is, therefore, not flexible enough as a design tool. Moreover, photographs are not suitable for visualizing major changes in the landscape.Photo montage on the screenA more efficient method of visualization is to manipulate both the plan and the photographs of the surrounding landscape by computer (colour photographs 1a to f). With a scanner, photographs are input pixel by pixel into digital memory. Video images can be digitized even more directly, with a video digitizer that "grabs" the image from a video camera or recorder (photograph 1c). Video images can also be mixed with the video signal of the graphics device (photograph 1d). Although both mixed and digitized images can be displayed on the graphics screen, mixed images can only be manipulated to a certain extent.In either case, the designer has realistic pictures of the background environment of his plan during the entire design process. If he uses software to generate the corresponding projection of the plan and place it in its exact location on the image of the environment, he can remove hidden parts and harmonize colours right on the screen.Realistic displays of natural elements can be obtained by copying existing trees and plants from digitized images of the landscape (photograph 1e). Computer programs can also change grabbed images considerably without any loss of quality. Because of the advantages, we decided to use grabbed images, rather than mixed video images.The hardwareTo develop the tool, we used a custom-made modular graphics system. The system has a video digitizer that grabs an image in real-time (25 frames/sec.) from a video recorder and enters it into its main image memory. This memory has 24 bitplanes, enabling the display of 16 million colours simultaneously. In addition, there is an overlay memory with 8 bitplanes (256 colours) for the display of computer-generated projections. Four images can be stored simultaneously in the image memory, but each image is displayed individually, with a resolution of 512 x 512 pixels (photograph 1f). The graphics system is connected to a host minicomputer (VAX 11 /750).The softwareAn existing computer program was used to generate the projections of the plan. This program, called VISTA (Visual Impact Simulation Technical Aid), was developed at the Strathclyde University (ABACUS group, U.K.) to visualize building designs. Our first step was to adapt the program to the hardware that was available at the Department of Computer Science of the University of Agriculture. During the research project it was rewritten and considerably enlarged. In addition, a program developed by the Department of Geodesy of the University of Agriculture was incorporated into the software, to compute the viewing parameters that correspond with the grabbed images of the environment. The software to manipulate the grabbed images was developed entirely within the Department of Computer Science.How to use the software1. Input a rough model of the plan (colour photographs 4a to c). A rough model is an acceptable representation of the plan, without any technical details, in which the natural elements are provisionally represented by symbols (e.g. the figure in photograph 5a).2. Take the video images of the existing landscape on the site and then grab and store them in computer memory. While taking the pictures, note the necessary data for the computation of the fitting projection: the position of the viewpoint, the date and the time, and, if possible, the focal length and direction of the camera (accurate values for the last two are difficult to determine, if a common video camera is used).3. When the viewpoint, the focal point, and the view angle of the corresponding image are known,input them. If they are not known, they can be determined with the software, which computes these values from a number of entered reference points. In practice, these computations are often cumbersome and not always possible. If that is the case, determine the viewing parameters Interactively, by comparing the computed projection with the grabbed image (photograph 2a). Add to the model reference objects that are clearly visible on the grabbed image. The correct values are those that produce a projection that matches the corresponding images of the reference objects in the grabbed picture (see colour photograph 7f; the tennis court is the reference object).4. Input the position of the light source, or input the latitude, the date, and the time from which the sun's position will be computed. To adjust the illumination of the projection according to the weather conditions of the grabbed-image, enter the colour, the intensity of the light source, and the ratio of direct to ambient light (if these parameters are not entered, default values will be used).5. The software will now compute the shaded projection with the hidden surfaces removed (colour photograph 2b).6. The computed projection is displayed in front of the grabbed image. Now the user can gradually change the projection colours until reaching a shading that harmonizes with the landscape (colour photograph 2d).7. Trace and cut from the projection the parts that will be hidden by surrounding elements (colour photograph 2c). If these elements have an irregular shape or are partly transparent (e.g. a row of trees), copy them from the original grabbed image and display them on top of the projection; then, to remove the parts of the old background that were copied together with those elements, filter as many pixels of a different colour as possible (see photographs 2e to h) until the projection reappears.8. Replace the projection of stylized natural elements with grabbed video images of the same sort. These images are fitted into the projection by pixel transformation (photographs 7a to d and 9e to h). Adjust the colours of the new, natural elements if necessary.9. Finally, store the resulting compositions in digital memory. For presentations, they can be recorded on videotape, film, photographs, or slides.ApplicationsThe system has been used to generate many perspective views. For example, colour photographs 3a and b show two bird's-eye views of a plan of the Markerwaard polder. Photographs 3c and d show views of a plan for forests near the city of Utrecht. Both plans were developed as entries for competitions. The projections in photographs 3e and f represent a first conception of the central attraction point of the Floriade 1992 (the city of Zoetermeer). All these views were made before the input and manipulation of video images were possible.The system's first compositions with video images were presented during the ceremonies to dedicate the new building that houses the Department of Computer Science and the Computer Centre of the University of Agriculture. A precise plan of the building had been entered into the system, which then generated a series of synthetic projections (colour photographs 6a to d). With perspective projections that had been fitted into video images of the site (photographs 8c, d, and a), several visual effects were shown, e.g. the changing in colour of the building and the addition of a small park in front (photograph 8g). The system also showed how the building would have looked at other sites (photographs 8f and h).In cooperation with the engineering firm of Grontmij, the visual impact of new forests in the open, western part of The Netherlands (Midden Delfland) was shown. Several series of bird's-eye views of this plan were generated, from a high altitude to eye level, the latter corresponding with video images of the site. The projections at eye level were drawn as wireline pictures onto the corresponding, grabbed video images. Then, the projected, stylized forests - represented by green cubes - were replaced by grabbed images of existing forests (colour photographs 9a to h and 10a to h).Many images were generated as part of a visual impact analysis done by Grontmij for a planned acoustic screen along the highway between Utrecht and Amsterdam. Construction of the acoustic screen would require chopping down the tall poplars along the road. With the software, the poplars on the digitized images were first covered with copied fragments of the sky. The acoustic screen and the elements that would become visible after removal of the trees were modeled and projected onto the images. These elements had been copied from video images taken from points with an unobstructed view. Subsequently, the copies were adjusted to match the perspective in the projections of the corresponding elements, showing the landscape as It would be after the trees had been chopped down and the acoustic screen had been constructed. In addition, the use of different materials was simulated by copying existing acoustic screens from digitized slides, thus enabling a visualization of different design concepts (colour photographs 7e to h and 11 to 15).This experiment revealed that the system can be used not only to visualize new elements in the landscape, but to simulate the removal of existing elements as well. In cooperation with the Organizing Committee of the 1992 Floriade, a series of slides was made for the presentation of the basic plan of the exhibition grounds. Besides several maps, each outlining a different aspect of the plan, a series of bird's-eye views was generated, one of which was mounted onto a digitized aerial photograph of the site. Finally, for a more natural look, fragments of digitized images were fitted Into a computer generated perspective projection (eye level) of the exhibition grounds (colour photographs 16a to h).Future developmentsWith this system, realistic images of future situations can be composed with far less effort than detailed hand-drawn images. Nevertheless, further automation of the manipulation process and a comprehensive library of digitized images of natural elements could considerably enhance its efficiency and usefulness. Also much work must be done to make the system into a full-fledged serviceable tool (e.g. transfer to faster hardware). The many positive reactions that have been recieved from colleagues and potential clients so far, however, are a good indication that further development is worthwhile.