Autonomously detecting and classifying traffic accident hotspots

Abstract

The number of road traffic fatalities has been steadily increasing since 2001 and is currently the eighth leading cause of death globally, with the loss of life of 1.2 million people each year according to the World Health Organization (WHO) [11]. In addition, the National Highway Traffic Safety Administration (NHTSA) reported that the number of deaths from traffic accidents in the USA increased by 7% from 2014 to 2015, rising to 35,092 fatalities [4]. Amid growing humanitarian concerns of so many injuries and fatalities worldwide, the Department of Transport issued a call to action encouraging the continuous research into different approaches that can improve the situation. As such, there are various research studies which are geared towards how in-vehicle systems can encourage drivers to adapt their driving behaviour and help to reduce the amount of both fatal and non-fatal traffic accidents. Typically, these systems aim to prevent a collision with an upcoming vehicle or pedestrian by providing warnings to drivers, and latest studies demonstrate promising evidence that these systems can indeed have significant positive effects [2, 9, 10]. However, the vast majority of studies have focused on simulation experiments [3, 8] and controlled lab experiments [6, 13]. We have recently contributed to this field by bringing an in-vehicle warning system into a field studying setting, utilising real world location analytics on traffic accident hotspots to generate in-vehicle warnings [7]. Going one step further, the benefit of ubiquitously detecting dangerous locations from data gathered by connected vehicles, and using these locations as a source for in-vehicle warnings,

A Machine Learning Approach for Classifying Road Accident Hotspots

“…Ryder and Wortmann [17] proposed an approach that detects and classifies accidentprone locations. They intended to warn drivers in real-time about imminent dangers on the road using a mobile application.…”

Role of machine learning approach for industrial internet of things (IIoT) in cloud environment-a systematic review

“…It includes the Adaboost application through which the driver's consciousness is assessed with the help of images. It helps prevent accidents by analysing the awareness levels of the driver [60]. Thus, it can be said that the application of IIoT systems in various smart city use cases such as intelligent energy, smart transportation, urban planning, and smart city characteristics help in enhancing the city functionaries.…”

A Review of Machine Learning and IoT in Smart Transportation

“…In a research about traffic accident hotspots and their automated detection and classification [32], the authors train an Inception Neural Network to help with the image classification. Analysis on road data obtained from the Swiss Road Authority (FEDRO) shows that there is a connection with the accident occurrence and the actual accident spot.…”

“…In the research [17], the drivers' consciousness is monitored with image data, processed by AdaBoost to prevent possible accidents. An Inception Neural Network is used in [32], to detect accident prone areas. Finally, the work of [24] to detect street elements could also support accident prevention.…”