Spatiotemporally Coherent Tensor Decompositions for the Analysis of Trajectory Data
Abstract
PURPOSE: Location acquisition technologies such as global positioning systems (GPS) sensors or telemetry devices generate abundant spatiotemporal measurements of movement of people, animals, and vehicles. The resultant data represent trajectories paths in space and time traversed by moving objects and can often be merged with additional information about the entities in motion from connected or external data sources (Zheng 2015). New data analysis frameworks may be able to uncover patterns of human behavior from the fused trajectory and contextual information. This data and new insights gained from novel analysis tools are potentially of great interest to the Army and the geospatial community. Methods for the analysis of collections of trajectories vary by objective. For instance, stochastic process models can be used for statistical inference of dependence among moving entities and uncertainty quantification of location from noisy or missing measurements (Scharf et al. 2016; Scharf et al. 2018); distance-based clustering methods using descriptive measures of spatiotemporal similarity provide a flexible suite of tools for searching for various inter-trajectory patterns (Zheng and Zhou 2011). Yet, these frameworks for the analysis of trajectory data, despite their many merits, provide no clear avenue for incorporating covariates or external data in order to understand relationships between movement patterns and other information.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 09, 2020
- Accession Number
- AD1103382
Entities
People
- Charlotte L. Ellison
- Trevor Ruiz