Incidents on expressways can cause congestion and impose
delay and other costs to the users as well as the society.
Timely detection of incidents is hence desirable and can best
be achieved through automated means. Although commercial
packages are available for detecting expressway incidents
automatically, their performance has been unsatisfactory when
implemented at a new site. For example, practically all
incidents on Singapore’s Central Expressway (CTE) were
detected by non-automatic means such as detection by the
Control Centre operators via surveillance cameras, as well
as telephone calls by the public and Police. One possible
reason is that most existing incident detection algorithms
were developed for a specific site, and the detection logic
does not work well when transferred to a new site with quite
different site characteristics and driver behaviour. In addition,
most of these algorithms were structured to receive traffic
inputs from conventional loop-detectors which are different
from the wide-area traffic data captured by the video-based
detector system along CTE. This research aims to develop
algorithms for the CTE which could accept wide-area traffic
inputs to provide efficient detection of incidents automatically.
A second objective is to develop algorithms that could achieve
good detection performance when transferred to other sites.

Authors:

Henry S L Fan (chenryf@ntu.edu.sg)
C L Mak (PA1686238@ntu.edu.sg)

Download the Full Text Here