Mining continuous activity patterns from animal trajectory data
The increasing availability of animal tracking data brings us opportunities and challenges to intuitively understand the mechanisms of animal activities. In this paper, we aim to discover animal movement patterns from animal trajectory data. In particular, we propose a notion of continuous activity pattern as the concise representation of underlying similar spatio-temporal movements, and develop an extension and refinement framework to discover the patterns. We first preprocess the trajectories into significant semantic locations with time property. Then, we apply a projection-based approach to generate candidate patterns and refine them to generate true patterns. A sequence graph structure and a simple and effective processing strategy is further developed to reduce the computational overhead. The proposed approaches are extensively validated on both real GPS datasets and large synthetic datasets.
Citation Information
Publication Year | 2014 |
---|---|
Title | Mining continuous activity patterns from animal trajectory data |
DOI | 10.1007/978-3-319-14717-8_19 |
Authors | Y. Wang, Ze Luo, Yan Baoping, John Y. Takekawa, Diann J. Prosser, Scott H. Newman |
Publication Type | Article |
Publication Subtype | Journal Article |
Series Title | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
Index ID | 70136278 |
Record Source | USGS Publications Warehouse |
USGS Organization | Patuxent Wildlife Research Center |