VISAR - VISUAL Analytics And Rendering

The overall goal of this project is to enable a seamless visual analysis of data involving 3D geometry, relational information and multivariate attributes. The key approach is to research new ways for achieving an unprecedented level of integration between realistic real-time rendering of 3D geometry and techniques from the field of information visualization and visual analytics. In a number of application scenarios, such as construction and inspection of large infrastructure, traffic visualization and archaeology, 3D models are used that capture the real world at a very high level of detail, and store a large number of additional properties for each represented object. In order to visually analyze such models it is necessary to provide a number of visualization tools that can deal with such heterogeneous data in a seamless and intuitive way. The project will research an integrated framework for visualization, navigation and interaction of both abstract and 3D geometric data. The research will be driven by a user-centered approach where successive prototypes will be evaluated in user studies. The tight integration of real-time rendering with state of the art information visualization will extend the applicability of visual analytics to problem classes that are tightly coupled to detailed realistic 3D models of the real world.

There is a wide variety of use cases involving complex heterogeneous data comprising 3D geometry and multivariate data. For instance, in the domain of tunnel maintenance it is essential to document cracks in the concrete surface. Each of these cracks is described by a 3D polyline and by multivariate attributes such as width, orientation, and moisture. Wet cracks running along and across the tunnel often indicate that a piece of concrete is prone to falling off. To find these critical patterns and adequately initiate repairs, experts need a simultaneous visualization of 3D geometric data and multivariate attributes.The main goal of VISAR Visual Analytics and Rendering is to explore new means for a tighter integration of realistic 3D real-time rendering and techniques from the field of information visualization in order to support interactive visual analysis of the described heterogeneous data. The combination of 3D real-time rendering and information visualization is becoming increasingly popular in fields such as urban planning, neuro science, or disaster management, to only name a few. However, obstacles inherent to combining 3D space and non-spatial representations are often ignored. An analyst finds an interesting outlier in a scatterplot, but may experience difficulties in locating its representation in 3D space. Further, no guidelines for visualization designers exist on how to build integrated solutions that offer efficient interactions to allow users the seamless exploration of both data domains.In this project, we developed a theoretical model describing the design space of how 3D real-time rendering views and information visualization can be integrated by using coordinated interactions on data, navigation, and the actual rendering. We further built a conceptual framework providing guidelines for visualization designers on how to develop integrated solutions, which foster an effective and intuitive visual analysis process of heterogeneous data.We applied our theoretical work to multiple use cases. In one application, lighting designers are confronted with a multitude of possible lighting setups to illuminate an office space involving contradicting goals, such as illumination policies for workplaces, energy consumption, and price. We combined an interactive global illumination application with a novel ranking visualization, which allowed users to identify optimal lighting solutions much quicker without relying on trial and error methods. In another use case, urban planners need to compare up to 30 buildings based on how they visually influence the cityscape in the context of historic lines of sight or characteristic viewpoints. We created different metrics to quantify visual influence, such as blocking of a landmark or sky occlusion including a fast method to gather visibility information from a 3D city model. A novel ranking view allows users to sort and filter candidate buildings with respect to the developed visibility metrics and a 3D view enables them to judge the spatial distribution of visual impact across the city. Besides our core research efforts, we could support other use cases involving heterogenous data. Furthermore, we achieved advancements in related domains such as visual parameter space analysis and system integration.