Innovative Techniques for Analyzing Cyclist Behaviour and Predicting Cyclist Safety

Cycling is an important mode of travel which offers powerful solutions to chronic traffic problems of congestion and emissions. Therefore, it is imperative for the road designers to have reliable information on the cyclist behaviour and characteristics to design safe and efficient cyclist facilities. The behaviour of the cyclist can be measured in different ways, including, cyclist crossing speed, cyclist violations, and cyclists- vehicle interactions. The essential focus in this thesis was on cyclists, especially their speed measurement and analyzing cyclist-vehicle interactions. In this thesis, a reliable video analysis technique to measure cyclist speed at signalized intersection was developed. This technique enables the automated observation of large volume of naturalistic cyclist movements in an accurate and resource-efficient manner. As for cyclist safety, relying solely on collision records to analyze cyclist safety is challenged by inherent limitations in collision data. These limitations can be quantitative and/or qualitative. Traffic conflict techniques have been used as a proactive and integrated approach to collision-based road safety analysis. However, traditional traffic conflict studies are mostly field-based studies and due to the subjectivity of field observers, errors were common when manually counting and deciding/judging whether a given traffic event is a conflict. In this thesis, interactions between motor vehicle and cyclists at signalized intersections were characterized using an objective conflict indicator; Post-Encroachment Time (PET). The thesis also produced a sizeable database of 806 hours of video data for cyclist movements. A total of 19,058 cyclists and 48,632 vehicles were observed within a total period of 57 days. The following contributions were achieved in this thesis: [i] video tracking system was improved from an open-source feature-based vehicle tracking system in order to track cyclist and vehicles and produce a trajectory database, [ii] different analysis methods were developed to measure cyclist crossing speed, [iii] automated measurement of cyclist crossing speed, [iv] investigation of the effect of different potential factors affecting cyclist crossing speed, [v] development and validation of an automated method to measure PET between cyclists and motor vehicles using video analysis techniques, [vi] different statistical techniques were investigated to utilize PET observations to measure a cyclist safety.