Fat Client visualization
Organizations involved: Fraunhofer (Germany)
The IQmulus Fat Client (IFC) is a desktop application for visualization which is integrated into the IQmulus infrastructure via an integrated web browsing component enabling control of the infrastructure, e.g. initiating workflows and accessing and visualizing results of workflows, using the main IQmulus UI. The visualization features are aligned to the requirements of the different showcases supported by the IFC and utilize modern OpenGL technology adapted to the visualization domain.
The IFC supports the expert user in decision making in a number of ways, which generally aim at increased performance of visual feedback for large data sets. This approach leads to faster analysis and decisions based on these data sets which increases the efficiency of the user and is especially relevant in time-critical situations. The most remarkable development in this context is the handling of large point clouds (i.e. interactive visualization and analysis of more than 1 billion points) which was not possible before IQmulus utilizing a single off-the-shelf PC.
For handling large point clouds a visualization pre-processing service running in the IQmulus infrastructure was developed, which processes one or multiple input point clouds with workflow specific attributes into a file-based acceleration structure. This acceleration structure is utilized by the IFC to identify the currently visible subset of the point cloud (depending on the current viewing position and orientation). The visible subset is streamed from the cloud storage to the local client and rendered taking a user definable point budget (maximum number of points to render) into account. The download of data is handled asynchronously, therefore navigation is possibly during retrieval of additional detail. Rendering is performed iteratively, i.e. the asynchronously downloaded data is integrated into the visualization as soon as it is locally available. To increase the perceptibility of details in the point cloud, an adapted ambient occlusion approach was integrated into the rendering.
Another important technique employed by the IFC is the use of deferred rendering, where the visible data is cached in off-screen buffers. This approach significantly increases the speed of data access in situations where the viewpoint is not changed. The approach enables instantaneous visible feedback to the user when inspecting attributes and derived values of shader-based local processing like color mapping.
To improve decision making in scenarios involving vector fields, the IFC provides dynamically tessellated glyphs generated directly on the GPU using tessellation shaders. This approach enables the user to navigate the data set and relieves him/her to manually adapt glyph density (i.e. the glyph density in screen space is kept constant independently of the current viewpoint). This allows for faster identification of relevant areas (usually large changes in the vector field).
Additionally, the IFC exposes parts of the programmable GPU pipeline to the user via a (rather) simple UI to specify shader programs (analysis algorithms) and dynamically adaptable parameters. This feature is highly interesting in scenarios where the characteristics of relevant data searched for is not known a priori or detection of this data requires manual tuning of search / analysis parameters. With this approach the user can change parameters of the current visualization algorithm interactively via automatically generated UI elements and get direct visual feedback on the results for GPU-local data.
An example of this visualization showing the 1.5 billion laser scan of Toulouse with detected trees (US2) is depicted in the Figure below. The visualization of vector fields for detection of landslides (LS4) is shown in the left part of the second Figure below, the user interface for shader-based analysis of raster data applied to Lidar full waveform analysis (MS5) is depicted in the right part of the figure.
FIGURE: 1.5 BILLION POINT CLOUD RENDERED INTERACTIVELY WITHOUT AMBIENT OCCLUSION (TOP) AND WITH AMBIENT OCCLUSION (BOTTOM). PLEASE NOT THE INCREASED VISIBILITY OF DETAILS WHEN USING AMBIENT OCCLUSION.
FIGURE: VECTOR FIELD VISUALIZATION USING GPU-SIDE TESSELLATED GLYPHS IN THE CONTEXT OF DETECTING LANDSLIDES (LEFT) AND USER INTERFACE FOR RASTER ANALYSIS USING GPU SHADERS IN THE CONTEX OF LIDAR FULL WAVEFORM ANALYSIS (RIGHT).