>>2008,ZHANG Yeting, Adaptive Visualization Methodology for 3D City Models
Abstract
With the quick development of three-dimensional (3D) geospatial data collection technologies, the virtual geographic environments become increasing elaborate with highly detailed and photorealistic 3D models of natural and man-made objects while maintaining a large volume of data. However the development of hardware and software is not sufficient to meet the real-time visualization requirements of such kind of complex models. Even the top-level computer cannot real-time display all these data. The development of hardware technology makes a wide variety of hardware terminals emerging, the performance of these terminals are different each other, even if they belong to the same type. In all kinds of 3D real-time visualization applications, how to output satisfied image according to different user requirments and hardware/software environment, which is named adaptive visualization, has become an international research point and the academic front. For on-line applications of large scale 3D city models, how to adaptively coordinate the rendering speed and quality according to factors such as data volume and hardware environment has been a critical issue of the applications of virtual geographic environments, CyberCity and digital earth.
There are several issues need be addressed to get the best visualization result in limited time. (1) How to optimize the speed of data loading and make the predicton of data loading time possible. (2) How to automatically adjust the system according to data and hardware condition changes so as to get the best image in limited time. (3) How to speed up the display stage without bad impact on adaptive mechanism. This paper firsly researched the technology content of the adaptive visualization of 3D city models. Then occlusion culling algorithm based on normal line grouping is brought out for speeding up the rendering efficient of complex building scene. To achieve time-critical adaptive visualization, most factors such as visiblity detecting, dynamic-loading of data, progressive rendering and hardware performance in all steps of data loading and display stages in 3D city models real-time visualization are considered together.The scheduling and rendering time of 3D city models are accurately estimated based on multivariate statistical theory. Visual perception theory and optimal choice theory are used to choose the model set that can achieve the best result image in limited time. The main contents of this research include:
(1) Difficulties analysis of 3D city models visualization. Based on the 2D map adaptive visualization research contents, the key technologies of the whole process of 3D city models adaptive visualization is analyzed and abstracted into four categories, which is adaptive data store and management, adaptive data loading speed, adaptive rendering anduser adaptive. The research progressions of the four categories are summarized and the technical difficulties are analyzed.
(2) Occlusion culling algorithm based on normal line grouping. For general complex building models, the reasons that make its rendering effect lower then other model is analyzed. As a result, a concept of visible cone of one component is defined to descript the visible area which considered the result of occlude by other components in this model. The visible cone is used to achieve quickly judge whether it is visible from current eye position in real-time rendering. The construct algorithm and the use condition of visible cone of component are discussed. For the rapid visualization application of complex model building that have clear directions, new algorithm “occlusion culling algorithm based on normal line grouping” is proposed based on the structure and data management attribute of complex building models. The main idea is: by taking use of the attribute that many nearby components have same direction and always become visible or invisible together, according to the main directions, component directions and nearby relations, the building will be split into several groups of components that with similar visible scope, their common direction is saved as the normal line of each component group. In real-time rendering, the normal line of each group can be used to judge the visibility of all components in it quickly and easily. Components group also can be used as occluder when occlusion culling is used. The compose application of normal line grouping and occlusion culling can further reduce the data volume that needed draw. This algorithm can effectively reduce the impact of the large number of occlusion between the components on the rendering efficient.
(3) Time-critical adaptive visualization method of 3D city models. According to online real-time visualization applications, a technical framework of time-critical adaptive visualization is proposed. The relevant factors in data loading stage and drawing stage that serious impacting the drawing frame-rate are analyzed. According to these factors, a quantitative time estimate model is established, which can accurately forecast the required drawing time in each frame according to costumed hardware conditions and data contents. It can provide the basis for adaptive adjustment of the data loading and drawing stages; the concept of visibility of a model was proposed and be used together with other existing importance factors in order to maintain the most important details as far as possible under time-critical condition. A LOD select algorithm for time-critical adaptive visualization is proposed, which use the results of rendering time estimate and importance value calculate and choose a suitable level for each model so that the total rendering time does not exceed the budget.
Based on the research contents in this paper, a prototype system is build up by using Visual C ++ 6.0. Several experimental data are tested on it, which verified the validity andthe feasibility of the methods.
Occlusion culling algorithm based on normal line grouping can effectively reduce the geometric data volume of complex building models that need to draw, which means the improvement of rendering efficiency. It can provide a relatively more budget time for time-critical adaptive visualization, Time-critical adaptive visualization method of 3D city models provided an effective drawing time regulation model, which can be a solution for the growing number of three-dimensional city model visualization applications once they are under the limited condition.
Key Words
3D City Models, visualization, adaptive, LOD, time-critical
PhD. Dissertation