An activity‐based framework for detecting human movement patterns in an urban environment. Issue 4 (26th March 2021)

• Record Type: Journal Article
• Authors: Hosseinpoor Milaghardan, Amin; Ali Abbaspour, Rahim; Claramunt, Christophe; Chehreghan, Alireza
• Is Part Of: Transactions in GIS. Volume 25:Issue 4(2021)
• Page Start: 1825
• Page End: 1848
• Publication Date: 2021-03-26
• Subjects: Geographic information systems -- Periodicals; 910.285
• DOI: 10.1111/tgis.12749
• Languages: English
• ISSNs: 1361-1682

Abstract

Abstract: The continuous development of positioning technologies and computing solutions for the integration of large trajectory data sets offers many novel research opportunities. Among various research domains, the extraction of users' movement patterns is an important issue that is yet to be addressed. While many previous studies have analyzed human and animal movements from a predominantly geometrical point of view, additional semantics are still required to provide a better understanding of the patterns that emerge. User activity data provide important information resources to analyze and predict movement patterns in urban environments. This study introduces a computational framework that combines the geometric and activity‐based dimensions of human trajectories. First, the geometrical dimension considers a series of parameters (i.e., turning points, curvature, and self‐intersection) that are extracted by a convex‐hull algorithm and characterizes a given trajectory. Second, user activity transitions are modeled and then denote some recurrent patterns. Finally, geometric and activity patterns are integrated into a unified trajectory modeling framework. This favors the analysis of human movement patterns by taking into account the geometric and activity dimensions. The entire approach and framework have experimented with the LifeMap Korean trajectory data set commonly considered as a reference benchmark. The experiments showed how the integration of geometrical and activity‐based dimensions could provide a better understanding of the patterns and trends that emerge from a large trajectory data set.