Feature Tracking with Skeleton Graphs

Abstract

One way to visualize large time-dependent data sets, is by visualization of the evolution of features in these data. The process consists of four steps: feature extraction, feature tracking, event detection, and visualization. In earlier work, we described the execution of the tracking process by means of basic attributes like position and size, gathered in ellipsoid feature descriptions. Although these basic attributes are accurate and provide good tracking results, they provide little shape information. In other work, we presented a way to describe the shape of the features by skeleton attributes. In this paper, we investigate the role that the skeleton graphs can play in feature tracking and event detection. The extra shape information allows detection of certain events much more accurately, and also allows detection of new types of events: changes in the topology of the feature.