Objectives
Virtual reality can enormously benefit sustainable urban planning over a long period of time.  Effective visualization considerably relies on relevant and contemporary data, capabilities of interaction and analysis, and involvement of participants in the planning process. The goal of this work is to develop such visualization application for coastal cities by Lake Michigan in northern Indiana. To achieve this primary goal, a number of supporting objectives must be accomplished in this study. First of all, we will develop an effective 3-D data (in particular buildings) collection technique. Such data is essential for urban planning in terms of visibility analysis, cultural and natural heritage reservation, noise and pollution modeling, and tourism industry. With the proposed development, we will be able to provide 3-D building models over selected interest areas at the proper levels of details. As for the second objective, we will develop highly automated method to derive land cover maps for coastal zones (the area from shoreline, marsh, sand to the city land boundary and near shore areas). This effort will provide the most recent and accurate coastal zone information for Indiana, which can then be used along with historical coastal information for coastal change modeling and prediction. The third objective is to provide interest users with the capabilities to interact, modify, manipulate, and demonstrate the views such that different design scenarios and modeling results can be evaluated. In summary, the outcome of the proposed study will be virtual cities over the selected areas that will meet a variety of needs in urbanization modeling, water resource planning, fishery industry, and ecosystem protection, and ultimately benefit long term sustainable economic and environmental development in the coastal areas in northern Indiana.

Methodology
The proposed methodology consists of three tasks: 3-D building model generation, coastal mapping, and visualization. The Indiana 2005 DSM, DEM and ortho images will be used for this study. For building model generation, we propose a clustering approach based on dimensionality evaluation to determine the points on the same planar roof. To obtain a reliable solution, we will adapt the density based clustering approaches, where each data point is given a potential based on its neighboring points, and the point with the highest potential is considered as the cluster center. In the last step, building component boundary will be traced by using a modified convex hull algorithm and reconstructed by enforcing certain regularization constraints such as perpendicularity, parallelism and symmetry.

As for coastal mapping, we propose a fusion based classification solution. First we will create the DFM (digital feature model) by subtracting DEM from DSM. The resultant DFM and/or its derivatives will participate in image classification as additional ‘texture’ bands. Such texture bands can be standard deviation, slopes, curvature, co-occurrence matrix, or spectral decomposition such as Fourier or wavelet transforms. As different objects have different textures, the inclusion of these additional bands into image classification will enable us to further separate spectrally similar features. Finally, all the feature classes derived from classification will be vectorized by using modified Douglas-Peucker methods followed by a regularization process.

Both the 3-D mapping and 2-D mapping results from the above steps will be integrated into Google Earth for evaluation and presentation. We will customize Google Earth such that sitespecific social-economic and planning data can be combined with the basic geospatial data or information extracted through this study to build a virtual coastal city. In addition to standard Google Earth capabilities, the development and enhancement will also allow users to easily access relevant data providers and servers, and support on-line digitizing, editing and labeling interest features.

Rationale
Lake Michigan has about 110 miles of heavily urbanized and industrialized shoreline in the states of Illinois and Indiana. During the past 30 years, this region has become one of the top three most populated areas in the country. Such accelerated development demands a long-time and effective strategic planning for the limited land and water resources. Sustainable development is crucial for citizens, policy makers, and urban planners in this region. Realizing the above urgent need, the state Indiana plans to achieve, manage, and maintain sustainable development for its coastal cities along the shoreline of Lake Michigan.

Both planning and subsequent implementation highly rely on the availability of various social, economic, and geospatial data at different resolutions; the capabilities to integrate, access and visualize them effectively; and the support on interactive scenario analysis for decision makers and the public. However, at present, most of activities in planning and its implementation are based on 2-D geospatial data and 2-D landscapes, much of which is also actually not up-to-date. As a result, urban planners and their guests must often pay field trips to visit places that they can not interpret very well in offices. This substantially increases the planning cost and delays planning progress since many places by the coast are not easily accessible due to transportation difficulties. Moreover, current planning outcome is mostly presented in 2-D environment and lack of realistic perceptual effect. One may not be able to fully evaluate and appreciate the future city that they have designed or expected. Finally, citizens can only at present time be presented with 2-D paper map or display. It is impossible for them to be fully and actively involved in different stages of planning and its implementation, which hampers the citizen’s participation, involvement and awareness in public affairs, and may even retard the effective communication between the local government and its citizens.