The NCDOT manages the
second largest state-maintained highway network in the United States. To assist
the Department in making appropriate pavement maintenance decisions, this study proposes
to evaluate primary and secondary roadway pavement conditions using deep
learning, a state-of-the-art artificial intelligence technique.
In this study,
roadway surfaces will be videoed using a vehicle mounted camera. These video
footage and raw roadway surface images provided by the data collection vendor
will be labelled at a very fine resolution to develop high quality datasets for
training, testing, and validation, and then deep learning algorithms will be
developed to recognize roadway surface distresses in an automatic manner.
The
deliverables of this proposed study could provide the NCDOT with a second set
of distress ratings that can be used for validation purposes and bring
efficiency gains in the pavement survey process that would save the Operations
Program Management Unit considerable labor and time. Once these deliverables
are implemented, it is expected that pavement distresses can be automatically
recognized at a more accurate and cost-effective level, leading to a substantial
increase in the effectiveness of the pavement condition assessments carried out
by or on behalf of NCDOT.
