Learnings from the development of a traffic data fusion methodology

The number of sources of traffic data has increased substantially in the last decade. In addition to the traditional sources of data from traffic signal and freeway detectors, we now have probe vehicle data, Bluetooth and Automatic Number Plate Recognition among many. The ideal data set is a fusion of available data that maximises the strengths of each data set and avoids their weaknesses. The topics discussed in this paper include: 1. How the quality of input data is assessed and bad data filtered; 2. The strengths and weaknesses of different sources of traffic data; 3. Data fusion considerations; and 4. Desirable features of input data sources. Included is a discussion of how two different data sources can provide speed data for the same link at the same time where both data sources are accurate, but significantly different. Methods for resolving these differences in order to provide a single speed value to clients and downstream processes are discussed. The methodology also provides data to inform the selection of an optimal ‘mix’ of data sources to achieve the best coverage and quality outcomes with available budget by identifying the extent to which each input data source contributes to the resulting fused data set.


  • English

Media Info

  • Pagination: 13p
  • Monograph Title: AITPM National Traffic and Transport Conference, 15-18 August 2017, Melbourne

Subject/Index Terms

Filing Info

  • Accession Number: 01664033
  • Record Type: Publication
  • Source Agency: ARRB
  • Files: ITRD, ATRI
  • Created Date: Mar 22 2018 12:31PM