Article Preview
TopIntroduction
Spatial landscape design is typically the domain of landscape architects and urban designers. Both professions require considerable expertise to ensure that the usually fuzzy and complex landscape design process will lead to a successful landscape design. Spatial design evaluation is not only confined to persons with a design background but also involves policy decision makers. Sketch plans are a well known medium for communication between stake-holders in the early stages of the landscape design process. Sketch plans intend to support rapid and fluent exchange of ideas on basic concepts of the plan. Even non-designers can participate in a sketch process because of the simplicity of the (traditional or digital) sketch tools. Developing a 2D sketch plan into a realistic 3D model is not conceivable in the early plan design stage and thus left to the designers later on in the plan making process. However, for spatial design evaluation a 3D model is indispensable (Lai et al., 2010; Tress & Tress, 2003; Yu et al., 2007). Other researchers (Al-Kodmany, 2002; Harris, 2001) have shown that laymen have severe problems interpreting 2D (sketch) plans. Moreover assessments on critical issues like cost, energy, sustainability are almost impossible based on sketch input.
Sketch input can be created in vector format or in pixel format using digital tools. Examples of vector-based systems that aim at automated creation of 3D models from 2D plans are 3D Cadastre, CommunityViz and the ArcScene module of ArcGIS. In 3D Cadastre (Stoter & Salzmann, 2003) a 2D plan is extended into a 3D plan by adding 3D polygon for 3D geo-objects to the 2D cadastral data. CommunityViz is a tookit for 3D planning built on top of ArcGIS (Kwartler et al., 2001). One of these tools is Scenario 3D that utilizes GIS features for 3D terrain creation and a Build-out wizard for populating the terrain with buildings. Areas are designated to land use types with specific density rules. According to these rules, points are placed on the 2D map which in turn can be replaced by 3D models of buildings to create a 3D plan visualization. In the standard ArcScene module of ArcGIS from ESRI, a semi-automatic creation of 3D models from 2D plans is provided. Based on simple automatic height extrusion of vector building footprint data in combination with manual façade texturing techniques more visual realism can be attained. Other 3D objects (e.g., trees, street furniture) are further created from existing point layers. An example of a pixel-based system is Visualscan (Borsboom-van Beurden et al., 2006). In this system so-called Land Use Icons act as a 3D stamp of 500 by 500 metres which represents the imaged landscape for a specific land use. The 2D plan area is represented as a 2D coloured raster with each colour referring to a specific land use. The Visualscan scene is created by linking the raster cell land-use to the accompanying 3D Land Use Icon.
Vector-based creation of 3D models from 2D plans has some serious limitations because the height of the objects is unknown and footprints of buildings often have erratic shapes. Although architectural scenes can be created for visualization from 2D vector data, height extrusion of building footprints does not represent the externalities and details, such as e.g., roofs, chapels, windows and foundations, typical of realistic building design and visualization. Moreover, the manual texturing operation needed for advanced realism is time-consuming. With the pixel-based system realistic 3D models can be created with a high level of detail at little effort, but land features like road network and landmarks are absent. A road network supersedes terrain configuration and thus is in conflict with the constitution of independent tiles. Landmarks like a tower are unique objects that cannot be included in a tile representation; otherwise, it would lose its uniqueness when applied in a larger area.