An Advanced Approach Toward Pedestrian Safety Analysis and Prediction: Advances in Multisensor Data Fusion and GIS Techniques

A substantial component of a pedestrian safety assessment is classifying and ranking pedestrian collision regions. This research focuses on two pedestrian safety levels that initially identified the zones accumulating a high density of pedestrian crashes, followed by the zonal assessment based on the crash severity. The framework of this research focuses on identifying crash hotspots with three separate methods and comparing results that allow recommendations for further development of pedestrian safety measures in the future. The application of pedestrian crash zone ranking using three methods, namely Kernel Density Estimation (KDE), Kriging Interpolation (KI), and Hotspot Analysis (HA), has been described within the framework for this study using the city of London's central and western ends as the study area. The crash data was obtained from the hidden data handles provided by the Transport for London (TfL) data extraction tool from the website and filtered by severity, type, and coordinates of the incidents. The detailed data files included the past 5 years' data from 2016 to 2020; the 2021 data files are still being published. A semi-variogram showing the correlation of input data was developed, which determines the criteria for clustering for the KI and KDE methods. The results obtained by the KDE method visually represent the areas with the density of accidents based on address-match input only. This study is an experimental approach that can help policymakers form pedestrian safety policies. Moreover, future automation processes integrated with GIS can reduce subjectivity in identifying the hotspot zones from KI and KDE maps.