The solution is for underground and cut-and-cover tunnels.
It provides robust and instantaneous detection of incidents, at the service
of supervisory checkpoints. Adaptive and self-learning, the system performance improves continuously and the maintenance is done automatically.
The solution is being deployed into the Toulon tunnel,
to take control of the real-time analysis over
a hundred video streams.
3.5 km long, with 2 tubes, this tunnel, connecting the A50 to the A57, is crossed every day by 30,000 vehicles, driving through under Toulon city center.
The solution is built on AI models pre-trained on various underground infrastructures
and several thousand incident sequences.
TunnelWatch improves continuously based
on the experience acquired by the algorithms
in the proper infrastructure as well as in all the other tunnels analyzed by TunnelWatch.
Trained by generalized and generalizable data, TunnelWatch is able to address new underground infrastructure while sharing the intelligence capitalized through the analysis of those already covered.
Thanks to smart masks system, TunnelWatch product make integrators and maintainers work easier -
it can continually self-configure in order to maintain
its performance over time.
Using deep learning helps the system to take into account
many parameters for every sequence, in order to improve detection.
The solution improves automatically overtime,
in order to reduce drastically false alarms.
TunnelWatch takes advantage of AI to recognize forms
and areas in tunnels, used to geolocate incidents and define
regular traffic flow direction.
If a camera moves –vibration, impact or wear- AI updates
automatically these informations.
Built to provide a reliable and understandable information
in every circumstances, and to facilitate set up and maintenance, Tunnelwatch graphical user interface enables a better comfort
for supervisors, and enables saving time
for administrators and integrators.