Maximal Centroidal Vortices in Triangulations. A Descriptive Proximity Framework in Analyzing Object Shapes

Abstract

This paper introduces a framework for approximating visual scene object shapes captured in sequences of video
frames. To do this, we consider the hyper-connectedness of image object shapes by extending the Smirnov proximity
measure to more than two sets. In this context, a shape is a finite, bounded planar region with a nonempty interior. The
framework for this work is encapsulated in descriptive frame recurrence diagrams, introduced here. These diagrams
offer a new approach in tracking the appearance and eventual disappearance of shapes in studying the persistence
of object shapes in visual scenes. This framework is ideally suited for a machine intelligence approach to tracking
the lifespans of visual scene structures captured in sequences of images in videos. A practical application of this
framework is given in terms of the analysis of vehicular traffic patterns.