DESIGN OF A PANORAMA MAP WITH PLAN OBLIQUE AND SPHERICAL PROJECTION

Parallel projections are often overlooked when designing a panorama, since most rendering packages use the central perspective projection as a default. The plan oblique projection, a particular parallel projection, has the interesting property of depicting the plan view of a terrain in correct shape. This paper presents the design and production of a panorama map using the plan oblique projection in the foreground, in combination with a spherical projection of the horizon in the background. This combination of two complementary projections allows the panorama to depict the plan view of an area of interest in an orthometrically correct way, and at the same time to convey a clear three-dimensional spatiality. The paper illustrates the different design and production steps, including the selection of data (SRTM digital elevation models, Geocover satellite images, etc.), the tool for rendering (an in-house development), and the techniques for color correction (Adobe Photoshop). MAPPING VISITOR CENTERS FOR NATURE CONSERVATION Visitor centers in protected areas are important reference points for visitors. They welcome the guests to the protected area and provide essential information for the understanding and preservation of nature. BirdLife Switzerland (URL 1), the national organization of BirdLife International (URL 2), aims at the conservation of birds, their habitats and global biodiversity. Augmenting the environmental awareness of the population is an essential step towards the achievement of this goal. BirdLife Switzerland therefore decided to produce a naturally colored map showing the visitor centers for nature conservation in Switzerland and abroad. The map’s goal is to incite people to visit the centers, and, more generally, explore nature in their immediate proximity. The map’s attractive and eye-catching design should therefore take the reader on a voyage, either a virtual one in his mind – or a real one on his sturdy hiking boots. To achieve the desired eye-catching effect, two rather unusual three-dimensional projections were combined with colorful satellite imagery. LOW-COST DATA FOR PANORAMA RENDERING Since the project only had a limited budget at its disposal, low-cost or free data had to be used. SRTM (Shuttle Radar Topography Mission) by NASA is a potentially very interesting digital elevation model (DEM) that is available for free (URL 3). The publicly available version contains data samples each 90 meters, but unfortunately contains data voids. These “holes” are concentrated on mountainous areas, and thus particularly visible on our panorama. To at least partially solve this problem, we used a patched version of the SRTM DEM available from Natural Graphics (URL 4). Natural Graphics fixed the holes by interpolating inward from the edges, while trying to preserve the edge slope. After interpolation, procedural fractal displacements were added to give the digital elevation model a more natural look. For texturing the panorama, we compared various freely available data sets, and finally chose the GeoCover data set. GeoCover is available for the whole globe at relatively high resolution (URL 5). It is a combination of three Landsat ETM+ bands, each sharpened with the panchromatic band. The GeoCover Landsat mosaics are delivered for free in a Universal Transverse Mercator (UTM) / World Geodetic System 1984 (WGS84) projection. GeoCover comes in two flavors, one named GeoCover 1990, available at 30-meter resolution, and JENNY: DESIGN OF A PANORAMA MAP WITH PLAN OBLIQUE AND SPHERICAL PROJECTION another one named GeoCover 2000 at 14.25-meter resolution. The GeoCover mosaics are in false colors and contain clouds in certain areas. We therefore used GeoCover 2000 as base data, with local patches from GeoCover 1990 data to cover cloudy areas. This pre-processing was carried out using Adobe Photoshop. We used land cover data to perform color corrections of the GeoCover Landsat mosaic. Unfortunately, there is currently no seamless data set available for the mapped area. For Switzerland we used the Arealstatistik data set available from the Swiss Federal Statistical Office (URL 6). The Corine land cover data set covers the rest of the mapped area and is available from the European Environment Agency (URL 7). Both data sets have a resolution of 100 meters. DESIGNING THE PANORAMA Even in the era of Google Earth’s interactive three-dimensional maps, static bird’s-eye views are still an attractive eye-catcher. Since BirdLife Switzerland is the publisher of the map, it seemed obvious to design a three-dimensional bird’s-eve view. We decided to use a standard northern orientation of the view to facilitate map reading. After some experimentation with the central perspective projection (Figure 1), it became clear that it is not possible to show all visitor centers on an easily readable map. The centers concentrate in the northern area of the map, which is more distant and therefore extremely foreshortened. Fig. 1. Oblique view of Switzerland with central perspective. The northern part suffers extreme foreshortening. Instead of the central perspective projection, we therefore used the plan oblique projection. The plan oblique projection shows mountains and hills in an oblique view, but displays the ground in an orthometrically correct way, i.e. the plan view is depicted without any distortion. Map readers therefore see the terrain in a familiar “map-like” shape, which facilitates map reading. In the past, the plan oblique projection has been sporadically used for terrain maps, for example by Kersten for his map of Switzerland, which was published in 1980 (Figure 2). JENNY: DESIGN OF A PANORAMA MAP WITH PLAN OBLIQUE AND SPHERICAL PROJECTION Fig. 2. Kersten 1980, “Die Schweiz aus der Vogelschau”, Kümmerly & Frey, 1:300,000. The plan oblique projection clearly depicts the terrain in three dimensions, but it does not convey an impression of spatial depth as strongly as the central perspective would. This lack is due to the missing foreshortening that would intensify the perceived three-dimensionality. An alternative method to add spatial depth to a panorama is the use of a curved horizon. An extreme example of this technique is Berann’s panorama of the alpine route through the dolomites, which (unevenly) curves the horizon by more than 90 degrees (Fig. 3). We achieved a similar, although substantially less striking effect, by applying an arc-shaped transformation to the image using Adobe Photoshop (It is recommended to use the “Upper Arc” transformation of Photoshop CS2 that is not altering the lower margin of the image). Figure 4 shows a prototype map that was produced at an early design stage, combining plan oblique projection and a spherical horizon. This example also uses artificial foreshortening that was added with Photoshop in the upper part of the image. JENNY: DESIGN OF A PANORAMA MAP WITH PLAN OBLIQUE AND SPHERICAL PROJECTION Fig. 3. H. C. Berann, 1958. 50 Jahre Dolomitenstraße. Fig. 4. Prototype map combining plan oblique projection and a spherically curved horizon. JENNY: DESIGN OF A PANORAMA MAP WITH PLAN OBLIQUE AND SPHERICAL PROJECTION RENDERING THE PANORAMA To the author’s knowledge, currently no commercially available terrain rendering software offers the plan oblique projection for panorama rendering. The described map was therefore produced with a prototypical inhouse development that is capable of draping geo-referenced raster images over the digital elevation model. Rendering occurred in multiple passes, i.e. views with shading only, the texture only, and the land cover only were produced independently. This approach allows for an exact control of the final colors of the map. It is to note that a 3D-renderer has to resample the texture image when draping it over a digital elevation model. Most rendering engines only offer bilinear or bicubic resampling for this purpose. For land cover data however, the software should also offer nearest neighbor resampling to avoid the creation of artificial intermediate classes. The rendering engine of the prototype software was extended in this way. Additionally, we also added support for the generation of so-called height maps, which code the elevation as grayscale values. The height maps were used for subsequent color corrections. Fig. 5. Separate rendering of a shading, the texture, the land cover data, and an height map. COLOR CORRECTIONS When working with GeoCover satellite imagery, the cartographer has to confront two problems. (1) The GeoCover satellite mosaic contains “false” colors that have to be corrected. (2) The GeoCover image is an amalgamation of a series of individual satellite images, and therefore shows locally variable tints that must be homogenized. Correcting false colors and homogenizing locally variable colors have been performed using Adobe Photoshop’s layer technique. Correcting the colors after projecting the satellite image to the third JENNY: DESIGN OF A PANORAMA MAP WITH PLAN OBLIQUE AND SPHERICAL PROJECTION dimension proofed to be more comfortable and effective than correcting them before the projection. This approach also offers finer control over the final product. The height map has been used for various color corrections, for example for separating the false-colored alpine meadows in purple from the constructed areas with a similar purple in lower areas. The land cover data has been used to further differentiate and correct color tints in the more detailed satellite image. For example, agricultural cropland was separated from urban areas. This approach allows for correcting the colors in the satellite image based on thematic land cover information, but preserves the granularity and structure of the original satellite image (Patterson 2002 and Patterson and Kelso 2004). Figure 6 and 7 show the final panorama after local retouches by the professional panorama painter A. Rohweder of Gecko Maps (URL 8). Fig. 6. The final bird’s-eye view