What street view imagery features favour driving? A copula model for driver distraction and driving performance

Abstract

Urban landscape plays a crucial role in reshaping the activity and mobility pattern of citizens. Studies have explored the relationships between the built environment, socio-economic, transport infrastructure, travel behaviour, and quality of life at different spatial scales. However, associations between the built environment, driver distraction, and driving performance at the micro-level are less studied. In this study, influences of different visual objects from drivers’ view and other possible factors on driver distraction and speed variation are investigated. Based on the street view imagery and image segmentation technique, proportions of visible objects including vegetation and road furniture within driver perspective can be estimated. Furthermore, vehicle kinematics in terms of longitudinal speed, longitudinal acceleration, and lateral acceleration can be measured from vehicle trajectory data. The Gaussian distributed copula model is used to jointly model the ratio of driver distraction and speed standard deviation. Results indicate that proportions of road, sky, and buildings in the drivers’ view significantly affect driver distraction ratio. In addition, speed standard deviation is associated with driver distraction ratio, proportions of sky and buildings, vehicle longitudinal and lateral acceleration, and driver age. Findings should shed light on enhancing urban design and planning by considering the effects of built environment attributes and drivers’ visual environment.