7.5 GIS Data Based Automatic High-Fidelity 3D Road Network Modeling GIS Data Based Automatic High-Fidelity 3D Road Network Modeling

30 road ~odels are widely used in many computer applications such as racing games and driving simulations_ However, almost all hl~h.-fideilly 3D road models were generated manually by professional artists at the expense of intensive labor. There are ver:J few eXlsllng methods for automatically generating 30 high-fidelity road networks, especially those existing in the real world. This paper pr.esents a novel approach thai can automatically produce 3D high-fidelity road networX models from real 20 road GIS data that mainly contain road. centertin~ in formation. The proposed method first builds parametric representations of the road cen~ertlnes through segm~ntallon and fitting . A basic set of civit engineering rutes (e.g ., cross stope , superelevation, grade) for road desl~n are then selected In orde.r to generat.e realistic road surfaces in compliance with these rules. While the proposed method ap~lIes to any types of roads, thiS paper mainly addresses automatic generation of compleK traffic interchanges and intersections which are the most sophisticated elements in the road networks. ' 1.0 INTRODUCTION Road networks are critical infrastructures of human civilization and probably the most important means of transportation. With advances in computing technologies, 20 and 3D road models have been employed in many applications, such as computer games and virtual environment construction. Roads are complex 3D structures and traditional road models were generated by professional artists manually using modeling software tools such as Maya and 3ds Max. This approach requires both highly specialized and sophisticated skills and massive manual labor. Procedural modeling based automatic road generation methods [1-6] create road models using specially designed computer algorithms or procedures and they can dramatically reduce or eliminate the amount of manual editing needed for road modeling. However, most existing procedural road modeling methods aimed at the visual effects of the generated roads , not the geometric or architectural fidelity that mostly determines the driving experience. Geographic information systems (GIS) are computer-based systems widely used to store, manipulate, display, and analyze geographic information in many fields. With the rapid development and widespread use of GIS techniques, vast GIS data are captured through various digital data collection methods such as remote sensing via cameras, digital scanners and Ll oAR. As one kind of them, road GIS data which record the information of the real road network in the best possible way have also been used in many applications and facilitate our lives greatly, e.g., the automotive navigation system. Since real road GIS data contain road network information which is indispensable for some applications, especially for transportation, homeland security and defense applications, it will significantly reduce both the time and lab cost if 3D road network can be modeled from road GIS data directly and automatically. However, there are very few existing methods for GIS based automatic 3D high-fidelity road networks generation. Most GIS based modeling work and software focus on buildings [7, 8], vegetation and rural landscape visual ization [9-11] rather than roads. Therefore, a method that can automatically produce 3D high-fidelity road network models from real road GIS data will greatly benefit numerous applications involving road networks. This paper addresses this problem by proposing a novel method which is used in an ongoing project to automatically generate 3D high-fidelity road network models from existing road GIS data in compliance with a set of selected civil engineering rules. The proposed method consists of several steps, including road GIS data preprocessing, road representation parameterization, civil engineering rules based road surface https://ntrs.nasa.gov/search.jsp?R=20110012094 2017-12-08T18:16:09+00:00Z