An Efficient Point Location Method for Visualization in Large Unstructured Grids

Visualization of data defined over large unstructured grids requires an efficient solution to the point location problem, since many visualization methods need values at arbitrary positions given in global coordinates. This paper presents a memory efficient, fast in-core solution to the problem using cell adjacency and a complete adaptive kd-tree based on the vertices. Since cell adjacency information is stored rather often (for example for fast ray-casting or streamline integration), the extra memory needed is very small compared to conventional solutions like octrees. The method is especially useful for highly non-uniform point distributions and extreme edge ratios in the cells. The data structure is tested on several large unstructured grids from computational fluid dynamics (CFD) simulations for industrial applications.