Tunnelwatch is an Automatic Incident Detection (AID) software which uses Deep Learning algorithms to improve the precision of real-time reported information in tunnels.
A tunnel is a particularly confined and therefore critical environment, always exposed to traffic challenges. Because of their constraints and responsabilities, tunnel operators are always looking for improved efficiency. Yet, the limitations of current AID solutions are glaring: tedious installation, lack of reliability, heavy maintenance, equipment dependence, etc.
Thanks to the latest AI technologies, Tunnelwatch provides a robust real-time incident detection solution for Supervisory Tunnel Center: it improves decision support while facilitating maintenance, for even more safety in tunnels.
Precision and reliability
Robust and intelligent detection, for more accurate and reliable data: use of Artificial Intelligence to eliminate false alarms.
Facilitated set-up and maintenance
Adaptive solution, with automated configuration: intelligent mask generation model for automated lane segmentation.
An open and autonomous system
Compatible with most cameras (no specific equipment constraints), making easier system maintainability and evolution.
Cross-tunnel learning: pooling of false alarm cases for shared intelligence that benefits all structures.
A simple and ergonomic interface dedicated to an efficient supervision for operators (integrator or supervisor)
Automated masking: during installation (automatic generation), as well as over time (adaptive maintenance)
Considering context (space, time, current incidents) to refine detections and consolidate correlated alarms
A single API, easily integrated into the tolling system - directly into the toll lane software or in back-office - without any equipment costs
of remote cloud-based tests
94 AID cameras
2 tubes x 3.5 km
60 000 vehicles per day
Toulon Tunnel Control and Safety Station
33% false alarms
20% incorrect incident classification
Installation in Toulon of a Tunnelwatch analyser for the 94 DAI camera video streams and a back-office server in the technical room of CET ESCOTA
Cyclope and ESCOTA called on renowned partners to properly evaluate the performance of the Tunnelwatch solution deployed in Toulon.
Lombardi SA was mandated to define a testing and performance analysis methodology and to carry out the execution of the project (night tests).
Our solution relies on AI models, pre-trained in underground structures and with several thousands of incident sequences
Thanks to Cyclope’s expertise in image processing, Deep Learning, and its ultra-powerful AI models for object recognition, Tunnelwatch eliminates almost all false alarms.TEST ON MY DATASET
Tunnelwatch continously improves through successive re-training phases that integrate all the situations encountered in the tunnels in which the solution is deployed. The exposure in all tunnels in which Tunnelwatch operates is thus shared to improve the overall performance of the system.TEST ON MY DATASET
Designed to meet the needs of tunnel supervisors, Tunnelwatch is the result of a collaboration between technology experts and tunnel operators. It generates qualified data, capable of providing relevant information aiding the operator through his decision-making
process, thus ensuring an even safer tunnel for users.
Tunnelwatch was designed to facilitate the work of integrators, by making interactions with the global supervision system and maintenance easier: API easy to integrate, self-configuring solution thanks to an automatic masking system, which automatically readapts in the event of a change.TEST ON MY DATASET
Detecting incidents throughout the highway network thanks to outdoor cameras: towards 100% virtualised security patrols
Providing an I2V data stream (infrastructure-to-vehicle) to help drive autonomous vehicles