Back to Top

Dublin Use Case


In the Dublin use case, the VaVeL prototype is used to identify automatically incidents in the streets of Dublin, analyzing the available data streams. At the same time, the effectiveness, the efficiency but also the usability and privacy preserving capabilities  of the developed framework will be evaluated in a real setting. In this use case VaVeL’s vision is to improve incident detection in the Traffic Control Room of Dublin City Council (DCC).

DCC, Intelligent Transportation System. Dublin City has been developed as one of Europe’s leading smart cities. The Roads and Traffic Department of Dublin City Council operates a ”smart” control center, aggregating live streaming data from sensors located around the greater Dublin city area to manage the City’s road network for the benefit of pedestrians, cyclists, motorists and public service and commercial vehicles. These sensors include:

  • SCATS vehicle counting sensors.
  • CCTV cameras.
  • Automatic Vehicle Location systems mounted on buses.
  • Weather and environmental sensors.

The department is currently pursuing the following strategic directions: (a) providing alternatives to car commuting, (b) developing, optimizing and maintaining the city’s road network, (c) managing on-street parking, and (d) improving the city’s environment.

 The Dublin Use case will study the following types of incidents:

  • Improve Traffic Detection. Currently the incident detection is performed by having operators manually monitor TV monitors and raw traffic data. Traffic Management. In VaVeL, the Dublin City Council plans to investigate the area of automated incident detection using the CCTV network, to alert the Traffic Management Centre of events that have just occurred. This would result in a faster and more efficient response to incidents in the Greater Dublin Area which in turn could reduce traffic congestion, aiding the economy and the environment. On a technical level, the system will leverage a massive array of multi-modal data, including static traffic sensors, GPS sensors on buses, CCTV video data, for applying a class of incident detection algorithms. To this end, video data that capture the scene for different incident types will be used in training the algorithms’ incident detection capacity. Further on, the technical innovation will move into combining video data and the other data sources relevant to a particular incident class.
  • Provide Open Video Data. The video data are currently only used by DCC operators. In the Dublin use case the expectation is opening up the data from the CCTV network as potential live feeds on the Open Data Portal of the City of Dublin.
  • Tracking Weather and Environmental Factors. Dublin City Council has an extensive network of environmental sensors including rainfall gauges, noise sensors and air pollution sensors. Currently these data are recorded for offline analysis. Even though, combining this rich information with other sources such as video streams and traffic volumes to detect or alert of potential incidents could result in an accurate view of the status of the city. This single pane of glass view would allow a city operator to evaluate the city as a whole very quickly. The system should leverage a variety of data sources, including environmental sensors, traffic flow information, and, potential incident information. In this respect, leveraging the outcome of incident detection work stream as a data source should be considered.

We are going to evaluate the proposed prototype at DCC control room for 6 months. The system will be used and evaluated by DCC operators. A weekly feedback loop will be established from the beginning of the pilot in order for the consortium to provide with a bi-weekly prototype update. A user interface will be established where DCC operators can mark incidents as True Positives (TP) and False Positives (FP). Unidentified events from the system will be marked as False Negatives (FN). The consortium will take advantage of reports provided by the authorities (traffic reports, weather reports). These documents will be exploited as ground truth.

The prototype will be evaluated in terms of the previously defined statistical measures TP, FP and FN. In addition, user satisfaction of DCC operators will be recorded. A clear situation comparison (with vs without VaVeL) will be provided in order to estimate the benefits of the system provided. The performance of the anonymization will be evaluated by labeling the video data. Furthermore, we will show that the video data is still suitable for further analysis. This will be demonstrated by showing that the anonymization step does not negatively impact incident detection algorithms.