While the techniques described above provide high accuracy halo extraction tools, they are limited to single-time, 3-dimensional slices through the overall data set. We estimated that extracting halos from particle positions by this method will require tens of minutes on the cluster of 16 PCs. This is not substantial, but when used across a data set of a few thousand frames, the time to completion becomes prohibitive (several months). Therefore an adaptive time domain algorithm is conceived and is being developed that augments the pure spatial method.
From an initial halo map derived from the spatial method, succeeding frames are used only to adaptively track identified objects. A centroid for each halo is recomputed for each frame and the population of particles is updated. This technique is very fast and requires less than a minute per frame to compute. After on the order of a hundred frames, a full spatial particle reduction is performed to rederive halos at that time step. These are correlated with the projected objects from the temporal method. New objects are identified as a result and a partial temporal roll-back sequence is then performed (going backward in simulated time) for those new halos, thus tracking them to their genesis.