Virtual City Maker And Virtual Navigator: A Modelling And Visualisation Solution For Mobile 3D Virtual Cities Christos Gatzidis 1 , Fotis Liarokapis 2 , Vesna Brujic-Okretic 1 , Stuart Baker 3 giCentre, Department of Information Science, City University, London, EC1V 0HB, UK {c.gatzidis, vesna}@soi.city.ac.uk 1 Faculty of Engineering and Computing, Department of Creative Computing, Coventry, CV1 5FB, UK F.Liarokapis@coventry.ac.uk 2 Alcatel-Lucent Telecom Limited, Christchurch Way, Greenwich, London SE10 0AG Stuart.e.Baker@alcatel-lucent.co.uk 3 ABSTRACT This paper presents a complete procedural 3D modelling solution for mobile devices, called Virtual City Maker, based on scripting algorithms allowing for both the automatic and also semi-automatic creation of photorealistic quality virtual urban content. The input data used include the combination of aerial images, GIS data, 2D ground maps and terrestrial photographs grouped with a user-friendly customized interface which permits for the automatic and interactive generation of large-scale, accurate, georeferenced and fully- textured 3D virtual city content. This content can be specially optimized for use with mobile devices but also with navigational tasks in mind. Moreover, a user-centred mobile virtual reality (VR) visualisation and interaction tool operating on PDAs (Personal Digital Assistants) for pedestrian navigation is also discussed. This engine supports the import and display of various navigational file formats (2D and 3D) and also includes a comprehensive front-end user-friendly graphical user interface providing immersive virtual 3D navigation to a wide range of users. KEY WORDS Applications, 3D urban modelling, mobile navigation 1. INTRODUCTION Today there is a growing need for computer-based, photorealistic visualizations of 3D urban environments in many areas including environmental planning, engineering, telecommunications, architecture, gaming, 3D city information systems and even homeland security. Therefore, the procedural modelling of virtual cities in 3D is a topic not only computer graphics research but also other fields such as GIS or photogrammetry have focused on for a number of years. Different approaches have been favored with the two main ones being a fully-automatic or a semi-automatic one. While the ideal 3D urban modelling system would be a totally automatic one, there are a number of obvious advantages to semi-automatic or user-assisted approaches where essentially an automatic system employs some limited operator input for guidance in a range of the tasks involved in order to offer better control on the resulting scene. Furthermore, the user-assisted (or interactive) tools can also be utilized as a more intelligent editing application for the preliminary model. For example, a popular approach is to offer suggestions to an automatic urban modelling system which then in turn concludes the modelling task at hand on its own. This can lead to a more efficient method for modeling complex structures. A more conventional approach to interactive modeling is to provide the user with a set of generic models that are then adjusted (using usually image data) altering the model and the viewing parameters [1]. In this approach, the system provides geometric computations but the drawback is that substantial time and effort are required from the user. Some other more recent approaches include the combination of user input with automatic processing introduced at various points with a different degree of control over the result. Offering an approximate building location to extract a building is one of the approaches suggested [2] that can lead to results although there is the important disadvantage of producing a final model very reliant on the automatic analysis. Other semi- automatic urban modelling tools have been described, some including methods for duplicating and/or cloning model meshes that are similar to others [3]. Yet another approach suggests an automatic system constructing topological relations amongst 3D roof points collected by a user for each roof. This method [4] results in a tool that can tackle easily several types of complex roofs. An additional notable interactive urban modelling approach involves the system handling complex building structures by means of constructive solid geometry [5]. This system also uses an image correlation method to fit a primitive to the image. It should be noted however that this last method can be very costly on processing power when modelling more complex urban sites. One of the areas where 3D urban models are also in need today, apart from the ones mentioned in the beginning of this section, is for urban navigation using mobile devices. This can range to a number of sub- applications including car and pedestrian navigation systems, location-based services and others. It appears that while the methods for automatic and semi-automatic creation of 3D urban models today do cover a lot of ground both in terms of concepts and also in delivery of content, they fail to take into account the considerable limitations and challenges mobile devices have to offer. Constraints in terms of graphics memory, processing power and memory cards render the output of most of these systems unusable on