Title
A Crowd Sensing Approach to Video Classification of Traffic Accident Hotspots.
Abstract
Despite various initiatives over the recent years, the number of traffic accidents has been steadily increasing and has reached over 1.2 million fatalities per year world wide. Recent research has highlighted the positive effects that come from educating drivers about accident hotspots, for example, through in-vehicle warnings of upcoming dangerous areas. Further, it has been shown that there exists a spatial correlation between to locations of heavy braking events and historical accidents. This indicates that emerging accident hotspots can be identified from a high rate of heavy braking, and countermeasures deployed in order to prevent accidents before they appear. In order to contextualize and classify historic accident hotspots and locations of current dangerous driving maneuvers, the research at hand introduces a crowd sensing system collecting vehicle and video data. This system was tested in a naturalistic driving study of 40 vehicles for two months, collecting over 140,000 km of driving data and 36,000 videos of various traffic situations. The exploratory results show that through applying data mining approaches it is possible to describe these situations and determine information regarding the involved traffic participants, main causes and location features. This enables accurate insights into the road network, and can help inform both drivers and authorities.
Year
Venue
Field
2018
MLDM
Countermeasure,Dangerous driving,Sensing system,Computer science,Hotspot (geology),Transport engineering,Traffic accident,Artificial intelligence,Machine learning
DocType
Citations 
PageRank 
Conference
0
0.34
References 
Authors
4
4
Name
Order
Citations
PageRank
Bernhard Gahr172.47
Benjamin Ryder262.73
André Dahlinger373.08
Felix Wortmann443.41