Trajectory Based Activity Discovery

Abstract

This paper proposes a framework to discover activities in an unsupervised manner, and add semantics with minimal supervision. The framework uses basic trajectory information as input and goes up to video interpretation. The work reduces the gap between low-level information and semantic interpretation, building an intermediate layer composed of Primitive Events. The proposed representation for primitive events aims at capturing small meaningful motions over the scene with the advantage of being learnt in an unsupervised manner. We propose the discovery of an activity using these Primitive Events as the main descriptors. The activity discovery is done using only real tracking data. Semantics are added to the discovered activities and the recognition of activities (e.g., "Cooking", "Eating") can be automatically done with new datasets. Finally we validate the descriptors by discovering and recognizing activities in a home care application dataset.

A fully online and unsupervised system for large and high-density area surveillance

“…In addition, researchers have also proposed some methods to detect abnormal activity in an unsupervised way [Hu et al 2007;Pusiol et al 2010;Fu et al 2005;Junejo and Foroosh 2007;Junejo et al 2004;Makris and Ellis 2003]. These methods are usually based on trajectory analysis with the help of tracking.…”

A novel laser-based system: Fully online detection of abnormal activity via an unsupervised method

“…On the other hand, researchers also propose some methods [7], [8], [9], [10], [11], [12] to detect abnormal activity with the unsupervised way. These methods are usually based on the trajectories analysis with the help of tracking.…”

Context-based trajectory descriptor for human activity profiling

“…The work of [3] classify the areas of interest as entry/exit zones, junctions, intersections, and stop areas, which are defined by trajectories characteristics. In [1] trajectory slow points define individual topologies which are combined to form the general topology. After that, they segment trajectories by topology affinity building a descriptor composed by primitive events.…”

“…Indeed, the research community has been focusing on solving technical problems associated to multi-tracking techniques and on encoding trajectorybased features to detect individual atomic actions. Some approaches combine scene and object features with trajectorybased descriptors to detect event primitives [1], while others aggregate interactions measures and cues to analyze small groups of pedestrian [2]. However, none of them explore the integration of scene objects with individual related features to form action profiles.…”