Authors: Sean Ahearn*, Hunter College - City University, Somayeh Dodge, Department of Geography, Environment and Society, University of Minnesota, Twin Cities, Minneapolis, MN, USA
Topics: Geographic Information Science and Systems, Quantitative Methods
Keywords: trajectory, segmentation, multi-frequency, clustering, classification
Session Type: Paper
Start / End Time: 8:00 AM / 9:40 AM
Room: Grand Ballroom A, Astor, 2nd Floor
Presentation File: No File Uploaded
Segmentation of trajectories is key to many movement analytics. The quality and quantity of GPS trajectories has increased to the point where automated analysis is the only feasible way to characterize and quantify movement patterns. A key issue is the need to train analytical algorithms on each new data set because model parameters vary with individuals, context and domains. Additionally, movement occurs over different spatial and temporal scales and thus requires an approach which can detect potential behaviors across these scales. The second challenge is the clustering of similar segments into categories for classification. This talk will discuss a new approach to automated segmentation of trajectories that has two key characteristics: 1) it doesn’t require parameterization and 2) it can segment a trajectory over multiple scales through its recursive application. The segments derived through this process will be clustered using similarity analysis as a precursor to classification into behavioral types.