Multi-sensor pedestrian position and attention tracking system
This master thesis describes the theoretical fundamentals, necessary design steps and implementation of an outdoor pedestrian location and attention tracking system for consumer studies. The approach introduced combines sensor data from multiple sources in a hybrid Particle/Kalman filter model. This system makes it possible to specify the position of test participants with 5 meter accuracy independent of time and walking distance in the evaluation area. Beginning with
an overview of already existing pedestrian positioning methods, the thesis describes the theoretical geodetic and mathematical fundamentals necessary for realizing the proposed localization system. After that, the used sensor devices and their measurement principles are introduced. The raw sensor readings are saved in a database specifically designed for geodetic measurements. The information stored is post-processed in iterative enhancement stages. The result of this data optimization is then visualized as a 3D model. These software components are described in the middle part of the thesis. In order to assess the performance of the system, two accuracy tests were conducted. The thesis concludes with the results of these evaluations and finally gives an outlook on further work.