Friday , 14 August 2020

MediaTunnel A real-time Traffic Incident Detection System

MediaTunnel, a real-time traffic incident detection system introduced a few months ago, was recently installed for the first time outside Europe – in the Reboucas tunnel of Rio de Janeiro, Brazil. A real-time traffic detection system enables efficient dispatching of emergency services and rescue teams, and facilitates in reducing the response time to traffic incidents and accidents. Results of MediaTunnel’s functioning in the city’s longest tunnel have been impressive.

Real-time Traffic Incident Detection is an important part of any modern traffic management system. Conventional traffic incident detection which is passive, detects an incident only after it has taken place. MediaTunnel from Paris-based Citilog Inc is a real-time traffic incident detection system for road tunnels which enables traffic operators to detect incidents by triggering an alarm within a few moments of occurrence of an incident. Operators can initiate appropriate action much before the incident can be detected by a conventional loop-based detection system, i.e., before the actual consequences of the incidents manifest. They can assess the type of rescue teams that need to be called in.

The system also facilitates official inquiries of incidents and allows for a deeper analysis of incidents such as observing the behaviour of road users before and after an incident. It also warns road users of traffic hazards due to the incident, thus reducing the risk of secondary accidents. Another advantage is it avoids the need for closing lanes during installation and maintenance work, closures which can cause quite a bit of losses to the economy. Also, it eliminates the problems of degradation of road surface and rapid wear and tear due to frequent digging. MediaTunnel has recently been installed in the Reboucas tunnel – the longest tunnel in Rio de Janeiro in Brazil linking Lagoa and Rio Comprido areas, two prominent work and residential zones in the city. Every day about 190,000 vehicles use the tunnel.

Pierre Champsavin, Business Development Engineer, Citilog Inc. Says, “The current algorithm used in the software of MediaTunnel is an evolution of the first incident detection system developed in the early 90s. The first consideration in the development of the software was to define what an incident was, in order to be able to detect it. The INRETS researchers figured the best way to know that there is an incident is to detect all superfluous vehicle stoppages which are not incidents. The second consideration was to develop an algorithm for comparing captured images, and to provide the quickest possible detection with the highest reliability. The system has to be accurate too in the face of light variations and differences of video stream quality.”

The system is capable of detecting a wide variety of incidents on normal traffic lanes as well as on road shoulders and access ramps, under traffic conditions varying from smooth-flowing to congestion to stop-and-go traffic. Incidents detected include stopped vehicles, traffic slow down, traffic congestion, movement of pedestrians on the road, loss of visibility and debris on the roads. It detects a stopped vehicle in congested traffic, distinguishing it from a vehicle which has temporarily stopped. It does this by comparing its stoppage time with vehicles that have stopped in traffic due to normal traffic congestion based on the fact that the stoppage time of the former extends for a much longer time. MediaTunnel gives an alert about impending traffic congestion in a lane when the average traffic flow speed falls below a preset threshold and the traffic density increases beyond a preset threshold at the same time. It also detects vehicles which drive abnormally slower than the average traffic speed by using dynamic traffic adaptation. Detection of pedestrians in the middle of smooth flowing traffic is a major safety hazard, and the system gives an alarm on such detection. An alarm is also triggered when a vehicle goes in the wrong direction. It detects debris on the road by classifying them as abnormally small still objects, and the fact that such objects are not found on the road during normal conditions. Driving conditions which create loss of visibility for drivers also trigger alarms.

The system is capable of detecting a wide variety of incidents on normal traffic lanes, as well as on road shoulders and access ramps, under traffic conditions varying from smooth-flowing to congestion to stop-and-go traffic. Incidents detected include stopped vehicles, traffic slow down, traffic congestion, movement of pedestrians on the road, loss of visibility and debris on the roads.

Rubens Rodrigues, Head – Traffic Management, CET Rio (the company which manages the Reboucas tunnel) says, “Earlier, when any incident happened we were not able to detect it for a long time, and traffic in the tunnel used to remain blocked. Therefore, we decided to install an automatic incident detection system to reduce our response time to traffic incidents and to minimise the impact of incidents on the traffic flow. We decided to install IP cameras in the tunnel as they provide a very good flexibility for installation and management of videos. So we needed an Incident Detection System with good capabilities with IP streams and MediaTunnel is good at this. The redundancy features of the system added another level of safety. Plus, the ability of MediaTunnel to say if the incident has occurred in fluid or congested traffic is important for traffic operators. We also found the number of traffic operators needed to monitor the traffic inside the tunnel has decreased, and most of the incident alarms given off by MediaTunnel have been reliable.”

The current algorithm used in the software of MediaTunnel is an evolution of the first incident detection system developed in the early 90s. The first consideration in the development of the software was to define what an incident was, in order to be able to detect it. The INRETS researchers figured the best way to know that there is an incident is to detect all superfluous vehicle stoppages which are not incidents. The second consideration was to develop an algorithm for comparing captured images, and to provide the quickest possible detection with the highest reliability. The system has to be accurate too in the face of light variations and differences of video stream quality. — Pierre Champsavin

Share with: