Analyzing People Flows

In order to analyze the motion data sets obtained by measuring people flows arsenal research develops sound statistical methods and machine learning approaches. Interesting research questions include:

• How to obtain the dwelling times of people in different spaces of an infrastructure (indoors and/or outdoors)?
• How to statistically infer the relative flows between different places of an infrastructure from mere directional counts in terms of origin-destination matrices?
• How to detect places in space-time where people frequently stop walking?
• How to determine main paths used by pedestrians?

The following image illustrates an example for the automatic detection of stopping people in a major train station based on video cameras and motion measurement by vision-based people tracking. Though the state of the art of people tracking is far from delivering perfect people trajectories, the measured data revealed interesting insights about places where people stop.

Zoom

The following image shows an example for locating and classifying mobile phone users’ most popular places (e.g. home or work) based on cell positioning technology (Cell-ID).

Zoom

Such semantic knowledge about a user’s context greatly improves Location Based Services (LBS). The collection of experimental data was conducted via a web service in cooperation with an Austrian mobile phone provider.