A framework for on-line detection of custom group movement patterns

Abstract

This chapter describes a lightweight approach for custom definition and detection of group patterns in a real-time analysis scenario, using a simple, yet flexible notion for groups of moving point objects (MPOs) travelling together. Groups are defined as sets of MPOs that are directly or transitively related to each other via freely definable binary relations. Group candidates are identified within a snapshot view of the MPOs at discrete time instances. By backtracking over previous snapshots, stable group compositions over previous time instances are identified and reported. We give insight about the used data structures and algorithms for the group candidate calculation and backtracking steps and illustrate the approach’s functionality with examples from a real data set.