CONNECT. TRANSFORM. AUTOMATE.
Road Safety Data Integration Using FMEBrandt Denham, B.Sc.Collision Data Supervisor / Spatial AnalystCity of Edmonton – Office of Traffic Safety
Introduction to the OTS
• The Office of Traffic Safety was established in 2006 • Supports national and provincial traffic safety targets
to help achieve reductions in traffic collisions and make streets safer for drivers and pedestrians
• OTS will reduce the prevalence of fatal, injury, and property damage collisions through the 4 E’s of traffic safety (engineering, education, enforcement, and evaluation) by improving data analysis and business intelligence, speed management, urban traffic safety engineering and road user behavior.
The Data
• The OTS has a vast amount of transportation and traffic safety related data at its disposal
• This data is stored in many different places, with many different owners in many different formats
• The ability to extract relevant, meaningful and accurate information in a timely manner is a MAJOR challenge
The Data
Collisions
Automated Enforcement Red light violations Speed violations
Neighborhoods
Police divisions
Traffic Surveys Traffic volume & speeds
Traffic Signals
Speed Limits
Road Network Road geometry Functional class
The Problem – Data Silos
Inadequate knowledge about the existence of various data and their availability
Lack of linkages with other databases resulting in duplicate data collection, processing and management
No standardized method for the specific identification of attributes across data sources
Lack of communication among stakeholders of important changes to the data
Lack of access to other data systems
Collisions
Traffic Signals
Roadway
Modified from the original picture published in http://blogs.sun.com/bblfish/entry/business_model_for_open_distributed
Traffic Surveys
The Goal: Complete Data Integration
QueryQueryCollisionsCollisions
AE ViolationsAE Violations
NeighborhoodsNeighborhoods
Police DivisionsPolice Divisions
Traffic SurveysTraffic Surveys
Traffic SignalsTraffic Signals
Speed LimitsSpeed Limits
Road NetworkRoad Network
Data Integration
• Data integration is achieved in 3 high level steps:
Step 1 – Create common geographic base layers
Step 2 – Clean and format datasets
Step 3 – Spatial Linking
Step 1 – Create Base Layers
• All datasets need a common geographic link, I refer these as ‘Base Layers’
• All OTS datasets are either located at an intersection or somewhere along a roadway segment or mid-block
• For the purpose OTS data, two base layers are needed • Intersection base layer • Mid-block base layer
Step 1 – Create Base Layers
• Unique reference points for intersections are created• Relatively easy using ‘Intersector’
transformer
• Unique road segment lines are created for mid-blocks• A lot of simplification of the road
network must be done
• Each point or line has a unique ID #
Step 1 – Problem Example
Problem: Cul-De-Sac roads with the exact
same name as the main road they branch off of
Why is it a problem? Two roads with the same name Creates an unwanted intersection
Solution? These Cul-De-Sac roads need to be
removed
172 Ave
172
Ave
Same Name
Step 1 – Problem Solution Example
Selects Cul-De-Sacs from the Road Network
Finds the neighboring
streets around each Cul-De-
Sac
Tests if any neighboring streets have
the same name
Snap roads back together after Cul-De-Sacs removed
Re-Merge roads with the
same name
Step 2 – Clean/Format Datasets
• Datasets come from various sources in various formats
• In order to integrate, all datasets must:
• Be spatially referenced• May require geo-coding if spatial reference is missing
• Be consistently formatted
Step 2 – Geocoding Problem Example
• Collision data from EPS does not come with a spatial reference, only a text location description• Ex) “Near McDonalds on 23 Ave”
• Data entry staff translate the location into an intersection or mid-block when entering the information into the OTS collision database
• Spatial reference still needs to be added• From the inception of OTS in 2006 until 2013, this was done
manually, adding points one-by-one
Step 2 – Geocoding Solution Example
• The base layers from Step 1 can be used to automate the manual geo-coding process This spawned another major
‘Automatic Geocoding’ FME project that was created to do exactly that
• Thousands of hours of time and money are saved
• Human error is eliminated
Over 37,000 points had been created manually. On average it takes 5 mins to enter one point. 37,000 x 5 mins… you get the point!
Step 3 – Spatial Linking
• Once you have achieved clean base layers and clean datasets, you can link the datasets to the base layers
• By spatially linking each dataset to the base layers, each dataset can be given the unique base layer IDs which can then be used to link one dataset to another
Step 3 – Spatial Linking Example
• In this example, two datasets have varying spatial accuracy but should be associated with the intersection of 100 Ave & 99 St
• A ‘NeighborFinder’ transformer can find the nearest base layer intersection to each dataset (you can also specify a max search distance)• They can then be moved to match the
spatial location of the base layer and both can gain the ID# attribute of the base layer
• After this is done, you can then link the Traffic Survey dataset with the Traffic Signal dataset based on the ID# from the Base Layer without actually needing the Base Layer
100 Ave
99
St
Base Layer Intersection(ID# 5457)
Traffic Signal
Traffic Survey Device
Step 3 – Spatial Linking
Utilizing the Results
• When all datasets are linked and accessible, we can turn the data into information and the information into knowledge
• The following example shows how integrated data was used to get a ‘full picture’ of data to do a comprehensive analysis of a particular problem location in Edmonton
2nd from Curb50%
3rd from Curb8%
Right Curb25%
Unknown17%
Collisions by Driving Lane (2012)
2012 data has less unknown traffic lanes so it may be a more accurate breakdown of the collisions by lane
The 2nd from curb lane is lane #3 (The right curb lane is not a through lane)
Chng. Lanes Impr.18% Fld. Yield
R.O.W.6%
Flwd. Too Closely
72%
Ran Off Road2%Struck Parked
Veh2%
Collisions by Cause (09-11)
1 2 3
Study area
The top 5 violators are all rental and cab companies.
0
5
10
15
20
25
Mon Tue Wed Thur Fri Sat Sun
Collisions by Day of Week (09-11)
Peak collision periods: Nov-Dec Christmas shopping Fri-Sat weekend shopping Mid afternoon shopping
23%
57%
20%
Average Monthly Speed Tickets Issued
Lane 3 (67)Lane 1 (77)
Lane 2 (195)
24%
40%
36%
Average Monthly Red Light Tickets Issued
Lane 3 (10) Lane 1 (6)
Lane 2 (11)
41.26%58.74%
Violator Registered Owner Postal Code
Within Edmonton
Outside of Edmonton
Conclusion
Integrated data builds a foundation for business intelligence We can’t manage what we can’t track
FME supplies the tools to take datasets in any format and make them consistent and linkable
The processes created in FME are repeatable and can be used to automate regular maintenance of integrated data
As an evidence-based organization, integrated traffic safety related data helps OTS and the City of Edmonton to make efficient and effective operational and strategic decisions
Mission of OTS
The City of Edmonton Office of Traffic Safety will reduce the prevalence of fatal, injury, and property damage collisions
through the 4 E’s of traffic safety (engineering, education, enforcement, and evaluation) by improving data analysis and business intelligence, speed management, urban traffic safety
engineering and road user behaviour
OTS Vision:
0Injuries and Fatalities
Thank You!
Questions?
For more information: Brandt Denham ([email protected]) City of Edmonton – Office of Traffic Safety (780)-495-9905