Road Tunnels Manual

You are here

Video Automatic Incident Detection (VAID)

Most road tunnels that have a manned control room are equipped with a Closed-Circuit TV (CCTV) system (refer to the page on CCTV system).The CCTV images are usually displayed on a video wall in the control centre.  The CCTV system allows operators to monitor the tunnel to identify incidents. It can often be difficult for the operator to monitor more than a few displays simultaneously because of the large quantity of images available and with other duties that the operator has to perform.

To ensure that incidents, such as slow moving or stationary traffic, are detected early, tunnel operators are increasingly installing Automatic Incident Detection (AID) systems.  In some countries, the use of such equipment is mandatory for certain types of tunnels.


Two main types of AID systems are available for use in road tunnels, Video-based and Radar based.  The video-based systems (V-AID) generally use fixed CCTV cameras to provide full coverage of the roads and walkways.  The video-based system provides useful time-stamped video images of incidents which can help in assessing why incidents occurred and who or what may be responsible for causing the incident.  Fixed camera V-AID systems are also available with Infrared cameras so that the system can provide video images in low light and smoke-filled conditions.  Radar based systems cover a greater area than one fixed camera, so require less infrastructure in the tunnel.  The output from a radar system is a grey-scale image that works in all environmental conditions in a tunnel.

V-AID systems are normally based on computer-based analysis of video image streams generated from cameras set up to view tunnel traffic. The video analysis compares each frame from the video camera with the previous frame and checks for differences outside what is considered to be ‘normal’ behaviour.  A number of algorithms are available from the VAID system which can detect a range of incidents, including:

  • stopped vehicles
  • smoke
  • vehicles moving in the wrong direction
  • speed reduction
  • slow vehicle
  • pedestrians
  • debris in the road tunnel
  • flames
  • entry into restricted zones
  • lost goods in the tunnel

Each type of event to be detected is associated with a false alarm frequency. Increasing the number of events to detect will therefore increase the occurrence of false alarms and thus reduce the reliability of the AID system.

It is therefore very important to limit the type of events to be detected to those that are strictly necessary in order to make the system as reliable as possible, to limit the frequency of false alarms and to reduce the cases of non-detection of events.

Since serious vehicle fires normally develop after traffic has stopped (e.g. following an accident), it follows that a 'stopped vehicle' alarm from an AID system can be expected to precede alarms triggered by other systems, such as temperature and smoke detectors. This early warning provided by AID systems provides time for tunnel operators to confirm the nature and location of the incident, and to allow more effective intervention. This may be through the choice of optimal ventilation configuration, prevention of secondary accidents through operational measures, rapid warning to motorists upstream of the incident, and tunnel closure.  It also gives operators the opportunity to call the emergency services, display warning or information messages on variable message signs, broadcast messages on public address and the radio voice break-in systems, call to breakdown lorries, advice to exit the tunnel, etc.

Video smoke detection systems are described in Section 6.3.3 "Currently Used Methods" of report 05.16.B 2006.

Fig. 1: Camera equipped with AID (photo Credit CETU)

Video-based AID systems can also provide real-time information on the traffic flow, volume and speed. They can record pictures at the origin of the incident (Fig. 1) and can interact with other systems such as the Supervisory Control and Data Acquisition (SCADA) system. Video-based AID systems normally include IP (Internet Protocol) cameras, a video image processing system processing images from one or several cameras. The video images may be fed back to monitors or computer displays in the control centre.  The V-AID cameras may also be monitored by the video management system, comprising of redundant servers, providing video and other functions (mass recording of video and incidents from the AID system, collecting and storing real-time traffic data and traffic events, interfacing with the tunnel SCADA system), network equipment and communications lines (optical fibres, coaxial and Unshielded Twisted Pair cables).

Recent technology in video processing allows the V-AID algorithms to be run on the cameras (or near the cameras), rather than on a central server.  This approach means that only video incident files of detected incidents are transmitted to the SCADA system when there is an incident, rather than constantly transmitting (streaming) real time video from all cameras.  This reduces the transmission load on the tunnel communications network


The design of AID systems in tunnels should be undertaken with due account of the following issues:

  • Types of incidents to be detected
  • Detection accuracy (i.e. minimisation of 'false negatives' in detection incidents)
  • Reducing false alarms (i.e. minimisation of 'false positives')
  • Location of existing cameras in the tunnel
  • Geometric features of the tunnel
  • Access for maintenance staff
  • Glare effect due to sunlight near cameras located in the vicinity of portals
  • Changes caused by the passage of vehicles in the tunnel (lights, occlusion by high vehicles)
  • Visibility in the tunnel caused by smoke
  • Change of lighting level in the tunnel (lighting on/off)
  • Reflections in the tunnel
  • In case of an AID system integrating IP video stream, the capacity of the existing IP network should be checked to ensure there is sufficient bandwidth available
  • Proximity to tunnel services such as ventilation fans, tunnel lightings, fire protection services etc.

The 2009 Routes/Roads article "Fire Detection Systems in Road Tunnels - Lessons Learnt from the International Research Project"concluded that "to deal with obstructions, most sensor manufacturers recommend two sensors covering the same area from different angles, such as from both directions within a tunnel". Multiple cameras may also be required for redundancy purposes, in case of camera failure. Typically, the camera fields of view are designed to overlap, such that failure of any one camera can be compensated through the images from neighbouring cameras.

Section IV.2.1. "Incident Detection Devices" of report 05.15.B 2004 suggests that camera locations can vary from 30 to 150 meters if they are used for automatic incident detection. More recent recommendations from the V-AID manufacturers is that the cameras should cover a distance of no more than twenty times its mounting height.  Therefore, if a camera is mounted at 5 m above the road then it should be able to provide reasonably accurate detection of incidents for a distance of 100m. However, as mentioned above, to ensure redundancy coverage in the event of a camera failure, then V-AID cameras should be installed every 50 to 80 m.

The performance of an AID system depends to a great extent on successful commissioning and calibration prior to deployment. Experience shows that the decision on detection time for the system has an important influence on the systems detection accuracy and the amount of false alarms registered. Experience from existing tunnel installations indicates that such commissioning and calibration can take several months under operation to undertake.

Reference sources

No reference sources found.