Assessment of the resilience of pedestrian roads based on image deep learning models

摘要

Currently, the evaluation of pedestrian paths is very time consuming. Additionally, disabled pedestrians do not tend tochange their routes, even if pedestrian conditions are poor, resulting in reduced convenience and safety. Therefore, it isimportant to identify and act on the statuses of pedestrian paths quickly. Therefore, this study aimed to identify andprocess the conditions of pedestrian paths quickly to achieve high resilience. A resilience triangle was calculatedaccording to the discrimination automation to analyse the corresponding values. Pedestrian path discriminationautomation applies convolutional neural networks and ‘you only look once’ analysis to identify the road surfaceconditions of walkways and the presence of obstacles. Quantitative analyses for the safety and economic problemsassociated with transportation vulnerabilities through discrimination algorithms using deep image learning werecarried out. As a result of the analyses, it was possible to determine the extent of damage with 94 accuracy if onlydamaged sidewalk photographs are captured. When this result was applied in Seoul, the benefits of improvingpedestrian paths were quantitatively calculated to be South Korean Won (KRW) 41.2 billion (1 KRW=US$0.00085). Thisstudy may secure pedestrian resilience and improve convenience in the current scenario of a rapidly ageing population.