Measuring the similarity of vector fields using global distributions

Abstract

Sensors such as video surveillance and weather monitoring systems record a significant amount of dynamic data which are represented by vector fields. We present a novel algorithm to measure the similarity of vector fields using global distributions that capture both vector field properties (e.g., vector orientation) and relational geometric information (e.g., the relative positions of two vectors in the field). We show that such global distributions are capable of distinguishing between vector fields of varying complexity and can be used to quantitatively compare similar fields. © 2008 Springer Berlin Heidelberg.