Categorisation of Computational Methods for the Extraction and Analysis of Vehicle Trajectory Data leading to an Increase in Road Safety

Abstract

In order to further reduce the number of accidents on our streets and at the same time to increase the efficiency of the available infrastructure, there is need to improve or even replace traditional traffic safety methods, which are based on accident data. In comparison, estimating accident risk is a preventive method, which analyses the single vehicle trajectories or the conflicts between interacting vehicles. To gain this information, vehicle trajectories can be computationally extracted from video recordings or from other data sources. To accelerate the analysis and to be able to handle the huge amount of video data generated, automated vehicle trajectory extraction is a promising method. Depending on the application, different levels of acquisition and analysis can be applied to achieve the necessary detection range, computational speed, realism or analysis complexity. This paper attempts to distinguish between those levels and presents several use-cases which require distinct levels in order to achieve their objective.