On the Impact of Classification Quality of Multiple Object Tracking Systems on Analysing the Path Choice Behaviour of Multimodal Traffic

Abstract

Multiple object tracking (MOT) systems are increasingly used for acquiring comprehensive motion data relevant to many fields of traffic research. Due to an increased interest in multi-modal traffic, the ability of such systems to distinguish between different object classes is a relevant quality criterion. This work discusses different aspects of classification quality and analyses the quality of two classifying systems at the same intersection in a comparative manner. Preliminary results show that the classification might be challenging for behaviour analysis and approaches to address the challenges are discussed.