Understanding traffic patterns and regular travellers using registration plate data
1. Understanding traffic patterns and
regular travellers using registration
plate data
Tom Cherrett, Fraser McLeod
Transportation Research Group
University of Southampton
2. Characteristics of commuter traffic
- 70% of commutes involve
travelling on local roads in a city
or town
- 74% in employment usually
work in a single work place
- Only 16% of commuters have
more than 1 route to work
3. Characteristics of commuter traffic
- 34% change route because of traffic
seen up ahead
- > income or education levels use
more than one route to work
- > the JT, the > the frequency of route
change
- Males change route more
- Older commuters make less route
changes
6. ANPR for understanding regulars
-Who should be arriving in the foreseeable future?
- How habitual are their behaviour patterns?
- Can we use ‘regulars’ as indicators or network state?
7. ANPR for vehicle analysis
25000
20000
15000
10000
5000
0
Number of observations
Year of registration
8. Research Questions
Using ANPR data:
• How habitual are vehicle arrival patterns?
• Can the arrival time variability of ‘regular ‘ vehicles be
used to gauge network performance?
• How does ‘churn’ affect the supply of ‘regular’ vehicles?
Could one use ‘regular’ vehicles as information carriers in
an ‘internet of cars’?
9. Dorset Test Site
• Dorchester to Weymouth
• 22 ANPR cameras
• 50 million observations over
12 months
Dorchester
Weymouth
11. Dorset ANPR data
• 50 million records and counting
• 118,200 records added each day
• Periods covered:
– 23/7/2012 to 12/11/2012
– 4/4/2013 onwards
• 76% of data have confidence level >=90%
6000000
5000000
4000000
3000000
2000000
1000000
0
COUNT(number)
0
32
35
38
41
44
47
50
53
56
59
62
65
68
71
74
77
80
83
86
89
92
95
98
number of plates in dorsetanpr
Confidence level
12. Average flow profile
(NB, weekdays only)
400
350
300
250
200
150
100
50
0
1 3 5 7 9 11 13 15 17 19 21 23
Average #plates recorded
Time of day
20. The problem of Churn
Turnover (‘Churn’) of regular vehicles occurs due to:
- Changes in vehicle ownership
- Changes in job status/working conditions
- Changes in home life
A traffic management system using the variability in arrival
rates of regular vehicles would need a constant update of
the ‘regular’ drivers
Churn was investigated by defining regular vehicles
(standard deviation of arrival time less than 10 minutes
based on more than 30 observations, 0630-0930)
21. Rolling analysis period
Churn investigated over rolling 4-month periods:
• Period 1 (P1) = 4/4/13 to 4/8/13
• Period 2 (P2) = 4/5/13 to 4/9/13
• Period 3 (P3) = 4/6/13 to 4/10/13
• Period 4 (P4) = 4/7/13 to 4/11/13
• Period 5 (P5) = 4/8/13 to 4/12/13
• Period 6 (P6) = 4/9/13 to 4/1/14
• Period 7 (P7) = 4/10/13 to 4/2/14
• Period 8 (P8) = 4/11/13 to 4/3/14
22. Rolling analysis period
30%
25%
20%
15%
10%
5%
0%
160
140
120
100
80
60
40
20
0
1 2 3 4 5 6 7 8
Percentage of vehicles
Number of vehicles
Number of 4-month periods in which vehicle was
a regular
23. Implications for network
management
- An additional method of monitoring issues on the
network
- Churn has implications for the supply of ‘regular’ vehicles
- How regular do vehicles need to be to identify potential
issues?
- Results suggest major roads during the morning
commute could be monitored in this way
- Interesting scope for live monitoring of different vehicle
types and CO2 footprints.