1 / 23

SATC 2014 7 July 2014

SATC 2014 7 July 2014. Uses of Automatic Number Plate Recognition (ANPR) in Traffic Management and Transport Modelling Alan Robinson Hatch Goba Alex van Niekerk SANRAL. Acknowledgements. Alex van Niekerk and SANRAL – content, ideas and use of the GFIP model access to ORT (ANPR ) data

ardith
Télécharger la présentation

SATC 2014 7 July 2014

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. SATC 20147 July 2014 Uses of Automatic Number Plate Recognition (ANPR) in Traffic Management and Transport Modelling Alan Robinson Hatch Goba Alex van Niekerk SANRAL

  2. Acknowledgements • Alex van Niekerk and SANRAL – content, ideas and use of the GFIP model access to ORT (ANPR) data • ETC - ANPR data • Nicholas Robinson – Data mining software development

  3. Uses of Automatic Number Plate Recognition (ANPR) in Traffic Management and Transport Modelling Contents • Data sets Traditional versus Big Data • GFIP and ANPR data • Overview of the ANPR data and outputs • Trend analysis, benchmarks • Traffic management • ANPR data uses in traffic modelling • Methodology for comparing ANPR data to model OD matrices • GFIP 2015 forecast validation • Conclusions and the way forward

  4. Traditional Data versus Big Data Traditional data; manual traffic data collection i.e. RSI, household surveys, trip diaries. • Pros: All data is obtained • Cons: Small samples so questionable accuracy Relatively high cost Big data; the utopia of traffic data using technology, includes: GPS tracking, Blue tooth tracking, GSM tracking, ANPR • Pros: Very large data samples • Cons: • Not all the data is there • It may include unwanted information, • It may need to be disaggregated using traditional data of questionable accuracy

  5. The GFIP Project • Comprises: • ±185km of Urban Freeway • 42 Open Road Toll Gantries • Sections with >150 000vpd • Systems on line in January 2011 • Tolling commenced 3 Dec 2013 • ORT systems records include: • Location (gantry number) • Number plate • Time / date • Vehicle classification • All other transaction data

  6. ANPR Data • Data collected from 4 weeks 24/7 during February 2014, • Raw data comprised 63 751 618 records • Number plates replaced by unique vehicle ID • Vehicle classes: • A1 – motorcycle • A2 – light vehicle <2.5m high • B – small heavy >2.5m high < 12m long • C – large heavy > 2.5m high > 12m long Raw ANPR/ORT data Sample data courtesy SANRAL/ETC Central Operations Centre

  7. ANPR Data Manipulation • Data Miner filters by day, time and vehicle class • Output: • Classified gantry counts in 15min time periods for each vehicle class • Number of records with no number plate (1%) • 32 652 058 trip records • Start and end gantry • Start and end time • Distance start to end gantry • Gantry to gantry speed • Gantry to gantry count matrices Data Mining Software Courtesy of NT Robinson

  8. ANPR Data Outputs • Detailed Traffic Counts per Gantry

  9. ANPR Data Outputs • Speeds between Gantries by time of day

  10. TRIP Data • Entries rolled up into trips

  11. ANPR Data • Gantry to Gantry counts (“OD”),

  12. Traffic Management Opportunities • Trend analysis of flows and travel times between gantries • Daily and weekly profiles • Seasonal profiles • Abnormal days • Set benchmarks • Real-time monitoring • Comparison to benchmarks • Exception reporting and alarms • CCTV verification • ITS driver information (VMS) • Response dispatch • Time series analysis • Trend analysis • Forecasts

  13. ANPR Data Uses in Traffic Modelling • Continuous traffic counts for assignments (calibration and validation) • 24/7 traffic counts • Standard ANPR cannot classify vehicles • Vehicle classification from ORT profilers • Journey times for volume-delay functions (validation only) • Speeds between gantries (over sections of freeway) by time of day • Can be related to the traffic volumes • Need single point speed and volume measurements for volume delay function calibration • Gantry to gantry Origin – Destination (partial for validation) • 24/7 continuous data by vehicle class • Relates to trips over the entire extent of the freeway network • Only partial trip related to the freeway ANPR data is comprehensive and accurate, we must relate this data to the model’s OD trip matrices

  14. The GFIP Traffic Model • Developed using SATURN • Based on provincial model • Comprises: • ±900 traffic zones • ± 22 000 road links • 6 user classes • 5 time periods • 2006 Base Year Calibrated to : • Land use data • HHS trip distribution • Journey time data from F/W and alternative routes • >600 automatic/manual traffic counts • Design Years 2011, 2015, 2020, 2025

  15. Comparison of 2014 ANPR Data to GFIP 2015 Design Year Forecasts • Comparison of traffic counts at gantry locations • Class A2: 14% high • Class B: 24% low • Class C: 22% low • All vehicles: 12.5% high Light Vehicles: 2014 Gantry Count vs 2015 Model Forecast Heavy Vehicles: 2014 Gantry Count vs 2015 Model Forecast

  16. Comparison of 2014 ANPR Data to GFIP 2015 Design Year Forecasts • Comparison of journey times and average speeds • Northbound OK • Southbound has a section where the model is too slow

  17. How to use Gantry-to-Gantry “OD” Data • Gantry-to-gantry “OD” id not a model OD • It is a count of the trips that only pass through the selected gantries and those along the route between them, i.e. A to B, B to C or A to C passing through B. • B to C would include from zones (4, 5, 6 and 7) to (12, 13 and 14) • In the overall model each approach would comprise trips from many of the model network zones

  18. Relating OD Information • Select Link through all gantries • Select Link through each gantry • Combination highlights cell groups only related to gantry combinations • A: only through A • AB: through A and B only • ABC: through A and B and C only • Etc. We can isolate model matrix cells that correspond to ANPR gantry-to-gantry counts

  19. The Workings • = Select link matrix through all gantry locations • = Select link matrix through all gantry locations • = Select link matrix through all gantry locations • = Trips from Gantry(a) to Gantry(b) only

  20. Examples of Gantry-to-Gantry X  X  X X  X  X X Network simplified for faster computer run times

  21. Comparison of Gantry-to-Gantry Trips to Model’s OD Matrix Sectors Comparison of Class A2 morning peak hour between: ANPR gantry-to-gantry counts and GFIP 2015 Model Forecasts

  22. Conclusions • All “Big Data” is good data but each set has limitations • ANPR data from toll gantries is comprehensive 98.9% sample of classified traffic movements on a freeway system. • The data is freeway focussed; not related to the complete trip. • We can use the data to validate various aspects of a model. • We devised a methodology to compare February 2014 data to the 2015 forecast from the 2006 GFIP traffic model • We are working on the functionality to calibrate and OD matrix using this data.

  23. For more information, please visit www.hatch.ca

More Related