Nearest neighbor search on moving object trajectories in secondo

Abstract

In the context of databases storing histories of movement (also called trajectories), we present two query processing operators to compute the k nearest neighbors of a moving query point within a set of moving points. Data moving points are represented as collections of point units (i.e., a time interval together with a linear movement function). The first operator, knearest, processes a stream of units arriving ordered by start time and returns the set of units representing the k nearest neighbors over time. It can be used to process a set of moving point candidates selected by other conditions. The second operator, knearestfilter, operates on a set of units indexed in an R-tree and uses some novel pruning techniques. It returns a set of candidates that can be further processed by knearest to obtain the final answer. These nearest neighbor algorithms are presented within Secondo, a complete DBMS environment for handling moving object histories. For example, candidates and final results can be visualized and animated at the user interface. © 2009 Springer Berlin Heidelberg.