Objektsensitive Verfolgung und Klassifikation von Fußgängern mit verteilten Multi-Sensor-Trägern

Abstract

State estimation of an unknown number of objects remains a challenging topic - despite the existence of theoretically bayes-optimal multi-object-filters - due to numerous assumptions in the modeling process. This thesis evaluates such filters in real multi-object-multi-sensor scenarios and proposes necessary extensions to existing models. The main application of the thesis is indoor pedestrian tracking.