GPU parallel algorithms for reporting movement behaviour patterns in spatiotemporal databases
Ignacio Valladares Cereceda · TDX (Tesis Doctorals en Xarxa) · 2013
In this thesis we treat and solve various problems related to movement pattern detection by designing and implementing parallel algorithms using the GPU. We first propose a GPU pipeline based algorithm to report the ’Popular places’ pattern. Then, we study the problem of reporting all subtrajectory clusters of a trajectory. To measure similarity between curves we choose the Frechet distance. Finally we solve the ’Flock pattern’. To this aim, we present two algorithms to solve two problems related with the ’Flock pattern’: finding the maximal sets of a family and intersecting two families of sets. The GPU parallel algorithms proposed to solve these two problems are later used for reporting flock patterns