Decoupling Mesh and Data Representations for Geo-spatial Data Visualization

Nowadays, data visualization plays an important role in geo-spatial data investigations, specifically in the context of visual exploration at interactive frame rates. Geo-spatial data typically include geographic information in form of a set of gridded or un-gridded points, where each point identifies a location on the earth’s surface. A variety of projection methods are in use to re-produce the location of a specific data point. For example, in contrast to the longitude/latitude projection system, the UTM (Universal Transverse Mercator) system provides a conformal projected location of the earth based on a non-linear scaling in both easting and northing. In this work, geo-spatial data such as terrain or ocean floor data are represented as 2D coordinates, e.g., longitude/latitude, and a number of associated attributes of measured information such as depth or height, backscattering, sub-bottom profile information, etc. Both main and graphical memory capacities are steadily increasing, but so are data sizes. Hence, current and future storage capacities still lack the ability to handle the targeted enormous data sizes with proper efficiency. To allow for interactive exploration, the number of rendered objects in most dynamic approaches is obligated to be below a certain complexity at run time. Fulfilling such goals requires several techniques to be integrated in our data exploration approach. View-dependency, LOD (level of detail) representations, and multi-resolution methods form the basis to tackle the objects redundancy, dynamic data updates, and simplification. Existing approaches can be classified into two main groups, the ones that operate on irregular triangular meshes and the ones that operate on regular grids. While irregular triangular meshes can provide better adaptivity, regular grids can be handled more efficiently. Our approach is to decouple mesh and data representations such that data management is performed efficiently on regular grids while mesh rendering is executed using a fixed adaptive triangular mesh in parameter plane. The triangular mesh is optimized with respect to the projected triangle sizes and shapes. During interaction we update the mesh by merely adjusting the vertices’ heights, which are queried from an underlying multi-resolution grid structure. The triangular mesh and the multi-resolution grid structure are precomputed and cached on the GPU. A tiling strategy is employed for the grid structure. We are able to generate very high frame rates using triangles with optimized shape in a