Incorporating Mobility Patterns in Pedestrian Quantity Estimation and Sensor Placement

Pedestrian quantity estimation receives increasing attention and has important applications, e.g. in location evaluation and risk analysis. In this work, we focus on pedestrian quantity estimation for event monitoring. We address the problem (1) how to estimate quantities for unmeasured locations, and (2) where to place a bounded number of sensors during different phases of a soccer match. Pedestrian movement is no random walk and therefore characteristic traffic patterns occur in the data. This work utilizes traffic pattern information and incorporates it in a Gaussian process regression based approach. The empirical analysis on real world data collected with Bluetooth tracking technology during a soccer event at Stade des Costières in Nîmes (France) demonstrates the benefits of our appro ach.