Arequipa Perú
VARIABLES INDICADOR VALOR FINAL ESCALA Variable Independiente
3. ANTECEDENTES INVESTIGATIVOS
In the final real world test the metrics are mapped back into the surrounding space, revealing visual exposure patterns for each visual property, as shown in Figure 3.21. In this example the visible number of targets increases with distance from the FOI, as does the Façade Area visible. These are measures of how much of a building structure is visible irrespective of distance, and as the observer moves away from the structure so more of it comes into view. However the Field of View and Perceived area maps show a decrease as the observation distance increases.
Maps may also be derived from the results, for example using flow direction functions to calculate in which direction the observer should move for a clearer view of the FOI. The direction to move for a more limited view may also be calculated, and it should be noted this is not an inverse of the clearer view map as explained in Section 3.5.4.
58
Direction to move for clearer view
=> clearer view => FOI Facade Area Field of View Clearness Targets Visible (%) Perceived Area FOI FOI
FOI FOI FOI
=> larger area =>
=> more visible => => wider angle => => larger area =>
Figure 3.21: Map of Visual Metrics
The dominance of an FOI within a region may be determined by ranking it according to one of the visibility metrics. This is achieved by totalling the calculations from many observation locations and linking back the results to each FOI. An example using the skyline metric within a university campus study region is shown in Figure 3.22. Here the total skyline percentage was calculated from every square metre in the study area, for each FOI. The skyline percentages from each observation point were then summed for each FOI, giving a single skyline value for the FOI for the entire region. In this example the FOIs with the highest rankings (i.e. most often on the skyline and not overshadowed) are the tallest structures (Figure 3.22 – FOI 3), or those with the largest floor area (Figure 3.22 – FOI 1), often located near the outskirts of the campus. The lowest rankings occur where a low-rise building is surrounded by much taller structures. This technique may be incorporated into the process of automatically identifying the most significant landmarks within an urban region.
59 Most Frequently on the Skyline
Least Frequently on the Skyline
1 2
3
4
5
Figure 3.22: FOIs ordered by how frequently they are on the Skyline when viewed from the surroundings The top 5 most frequent skyline FOIs are labelled, and also displayed in a 3D view generated from a LiDAR dataset.
3.8
Conclusion and Future Work
Current visibility modelling tools in GIS provide the functionality to estimate how much of the surrounding region is visible from an observation point. This is particularly useful for measuring the impact of planned developments on the surrounding region, or the feeling of openness in built-up environments (Fisher-Gewirtzman and Wagner 2003, Yang et al. 2007). Typically isovist models (Benedikt 1979) are used to describe the extent visible in urban environments, while viewsheds (Fisher 1991, Fisher 1995, Wang et al. 1996, De Floriani and Magillo 2003, Cooper 2005) have been applied in rural regions to calculate which portions of the countryside may be seen by an observer.
However only limited research exists for modelling the „visual exposure‟ (Llobera 2003) of a feature. This is an estimate of how much of a feature may be seen from the surrounding terrain by modelling its visual properties, such as the vertical field of view occupied, from every neighbouring location, such that an observer may be instructed in which direction to move to attain a clearer view of a landmark.
The research presented here uses 2.5D GIS datasets to model FOI profiles, such that façade areas may be calculated. This is achieved through an extended LOS model which considers the most prominent features in front of the FOI. Each FOI cell is considered as a vertical column between DTM and DSM models, and values for visible area are aggregated for all the visible
60 cells within a designated FOI boundary to generate object level summaries. These summaries are intended to quantify the visibility of FOIs, allowing future LBSs to add aspects of feature visibility into the array of searching and filtering tools available, and therefore to build a greater contextual understanding of the user‟s current situation. The visibility algorithms take into account the elevation profile of the FOI and the surrounding city, and are able to deduce if an FOI is out of sight, clearly visible, partially obscured, or on the skyline. FOIs may be ordered according to their perceived visible area by taking into account the building‟s visible extent, viewing distance and viewing angle which may be in turn used by LBSs to prioritise information delivery.
A number of trials were carried out on a synthetic DSM to establish that the algorithm‟s function, prior to running real world demonstrations. Adaptations were required so that concepts such as „on the skyline‟ and „clearness‟ indexes could be calculated successfully. The final model was able to produce results which reflect the experiences of the observer, quantitatively expressing the degree to which FOIs are visible, or obscured from view.
The algorithm is computationally intensive, which restricts its use on mobile devices directly. Two approaches which may be used to overcome this limitation are to pre-render the results for a set of FOIs in a study region (e.g. city, campus), or to provide a client-server architecture.
The pre-rendered approach works by first processing the visibility results from all possible locations within a designated study area to build a result cache. This requires a considerable computation facility but is largely aided by considering only those locations which are accessible to pedestrians outside of buildings, thereby reducing the number of observation points. In addition the algorithm may be run in parallel across a number of CPUs, as each target visibility test is independent. The FOI summaries do however require the results to be brought together on a single processor. Once the cached dataset is available it may be readily used in real time on current mobile platforms, which need only look up the appropriate results for a location. The main disadvantage of this approach is that any changes to the DSM or FOI database require reprocessing and cache distribution.
The client-server approach requires no advanced processing and provides real time server side support for a mobile client. Here the mobile application returns the location and orientation information to the server, which runs the visibility model in parallel across a number of CPUs to return the FOI visibility information in near real time. The main advantage of this method is that the DSM and FOI database may receive regular updates without the need to distribute this to clients. The main disadvantage is that sending data over wireless networks imposes an ongoing usage cost, which hopefully will reduce with the advent of WiMax (Ohrtman 2005, Patton et al.
2005) and other regional wireless network data solutions.
The processing time is largely dependent on the CPUs available, the number of FOIs in the database, and the resolution of the DSM. To improve the processing efficiency an additional step
61 may be added whereby the FOI database is filtered very quickly to determine which are visible, and therefore should be modelled in full. This is done by calculating the skyline polygon around the observer, and using this to find the intersecting FOI polygons. The skyline polygon requires that only a few thousand points are sampled, resulting in a significant performance increase in the overall algorithm, as the number of FOI targets which can be removed from the scan list will be significantly more than this.
There are a number of improvements which can be made to the algorithm including consideration of partial visibility through vegetation and methods to handle position uncertainty. These would both introduce a probability to the result, based on the shape and density of vegetation and accuracy assessment of the observer location. The latter should become less of an issue with the advent of improved tracking solutions to assist GPS in difficult environments, such as Inertial Measurement Units (Radoczky 2007).
Additional factors which influence the identification of FOIs, such as their use, shape, texture, colour, materials, and architectural design would be useful additional metrics to include in calculating a landmark dominance index, and to assist in FOI identification in urban scenes. Establishing the relationship between these metrics and the user‟s perceptions of landmarks will require user trials and should form the basis of an algorithm which can automatically establish, from any location, the most significant buildings, and those most easily identified.