Evaluating Remotely Sensed Images For Use In Inventorying Roadway Infrastructure Features

1
Evaluating Remotely Sensed Images For Use In
Inventorying Roadway Infrastructure Features
N C R S T
INFRASTRUCTURE
2
The Problem

• DOT use of spatial inventory data
• Planning
• Infrastructure Management
• Safety
• Traffic engineering
• Meet federal requirements (HPMS)
• Inventory of large systems costly
• e.g., 110,000 miles of road in Iowa
• Current Inventory Collection Methods
• Labor intensive
• Time consuming
• Disruptive to traffic
• Dangerous (workers located on/near roadways)

3
Research Objectives

• Investigate use of remotely sensed images for
  collection of roadway inventory features
• Evaluate level of resolution required for various
  inventory features
• Identify feasible features for future automation
• Make recommendations

4
Research Approach

• Identify common inventory features
• Identify existing data collection methods
• Extract inventory features from aerial photos
• Performance measures
• Feature identification
• Accuracy of linear measurements
• Positional accuracy
• Define resolution requirements
• Recommendations

5
Identify Common Inventory Features

• HPMS requirements
• Additional elements (Iowa DOT)

• Number of signals at intersections
• Number of stop signs at intersections
• Type of area road passes through (residential,
  commercial, etc)

• Number of business entrances
• Number of private entrances
• Railroad crossings
• Intersection through width

6
Required HPMS Physical Inventory Features

• Shoulder Type
• Shoulder Width
• Right and Left
• Number of Right/Left Turn Lanes
• Number of Signalized Intersections
• Number of Stop Intersections
• Number of Other Intersections

• Section Length
• Number of Through Lanes
• Surface/Pavement Type
• Lane Width
• Access Control
• Median Type
• Median Width
• Peak Parking

7
Data Collection Methods

• Manual (advantages/disadvantages)
• ? low cost
• ? visual inspection of road
• ? accurate distance measurement
• ? workers may be located on-road
• ? difficult to collect spatial (x,y)

• Video-log/photolog vans (advantages/disadvantages)
  
• ? rapid data collection
• ? permanent record
• ? difficult to collect spatial (x,y)
• ? may interfere with traffic stream

8
Data Collection Methods

• GPS (advantages/disadvantages)
• ? highly accurate (x,y,z)
• ? can record elevation
• ? time consuming
• ? workers may be located
• on-road

• Traditional surveying (advantages/disadvantages)
• ? highly accurate (x, y, z, distance)
• ? time consuming

9
Pilot Study
Ames, Iowa
10
Remote Sensing Datasets

• 2-inch dataset - Georeferenced
• 6-inch dataset - Orthorectified
• 2-foot dataset Orthorectified
• 1-meter dataset Orthorectified
• Simulated 1-meter Satellite Imagery
• not collected concurrently

11
Inventory Features Collected

• Crosswalks
• Left turn lanes
• Presence
• Length
• Width
• Stopbar
• Signal
• Structure
• Width
• On-street parking
• Presence
• Type
• Intersection design

• Through lanes
• Number of lanes
• Width
• Shoulder
• Presence, type
• Width
• Pedestrian islands
• Access
• Private
• Commercial/Industrial
• Adjacent land use
• Median
• Presence, type
• Width

12
Extraction
6 resolution image
13
Performance Measures

• Feature Identification
• Accuracy of Linear Measurements
• Positional accuracy

14
Feature Identification

• Number of features identified in aerial photos
  versus ground truth
• e.g. only 44 of the time can the number of
  through lanes be correctly identified (24-inch
  resolution)

15
Can the Feature be Identified?
16
Feature Identification
Observations is the number of features
tested. Differences by datasets indicate a
smaller available sample size
17
Linear Measurement Accuracy

• Linear features
• Measured in the field using handheld DMI
• Measured with 4 datasets
• Use of linear measurements
• Turn lane width -- intersection capacity analysis
• Driveway width access management
• Recommended accuracies from NCHRP Report 430
• Lane lengths within 3.28 feet ( 39.4 inches)
• Lane, median, and shoulder widths within 0.328
  feet ( 3.9 inches)

NCHRP Report 430 Improved Safety Information To
Support Highway Design
18
Thru Lane Width Error
Left Turn Lane Length Error
19
Linear Measurement Accuracy

• Lane width, turn lane length, and driveway width
  measurement relied heavily on pavement marking
• Expect less error with better identifiers (i.e.
  length of raised median)
• Accuracy required depends on application

20
Positional accuracy

• 50 GPS points were collected for comparison
• kinematic GPS
• 5mm to 10mm horizontal
• 4 cm vertical accuracy
• Compared to the same points located with the 4
  datasets
• Root mean square (RMS)

21
Results of RMSE and NSSDA Test (95th Percentile)
NSSDA Spatial accuracy test suggested by
National Standard for Spatial Data Accuracy
22
Cost Comparison

• Aerial photos
• Iowa DOT estimates 100/mile for images
  in-house costs to orthorectify
• 1.5 hours in-house to locate 55 features
• hour to measure turn left turn, 2 approach
  lanes, lane and median lengths, lane and median
  widths for 1 intersection (see davids)
• Field manual data collection
• 1 hour to measure and record turn lane and median
  lengths lane and median widths for 1
  intersection in field not including (see davids)

23
Cost Comparison

• GPS
• Cost 1500 for 55 points w/ kinematic GPS from
  consultant
• 24 person hours
• 10.5 hours for 1 person
• 3 hours processing
• All sites within 2 miles
• Videolog van
• 35/mile to collect
• How many miles can they collect per hour
  realistically, not including travel time to
  location
• Processing time
• Manual see Davids
• Costs for on-road data collection can increase
  significantly when sites are located at distances
  from data collectors and equipment
• 2 hours for Iowa DOT Mandli van to reach Iowa
  City from Ames, Iowa

24
Conclusions

• Majority of inventory features studied could be
  identified in the 2-inch, 6-inch, and 24-inch
  datasets
• Ability to identify features in 1-meter dataset
  is significantly reduced
• If identified, most features could be located
  spatially and measured
• Positional accuracy and linear measurement
  accuracy varied by dataset
• Acceptability of positional/linear measurement
  accuracies depends on application

25
RS for Inventorying of Roadway Features

• Advantages
• Rapid field data collection
• Multiple uses of data
• Data can be shared among state, local, etc.
• Do not need to return to the field for missed
  items
• Can collect most inventory elements (depending on
  resolution)
• Easily integrated with GIS
• Rapid in-house data collection

• Disadvantages
• Costly for initial collection of images
• (although multiple uses would decrease costs)
• Difficult to detect features such as signs

26
Applications

• Iowa DOTs linear referencing system (LRS)
• Identification of passing zones

27
Iowa DOTs LRS

• Iowa DOT is implementing a linear referencing
  system (LRS)
• Requires method to create accurate spatial
  representation of network for creation of datum
• Current accuracy requirements
• Anchor points (nodes) 3.28 feet
• Anchor sections (links) 6.89 feet
• Business data located - 32.81 feet

28
Data Collection Methods Tested by Consultants for
Iowa DOT

• Anchor points
• Kinematic GPS (reference dataset)
• Heads-up digitizing of 24-inch orthophotos for
  coordinates (meets)
• Heads-up digitizing of 6-inch orthophotos for
  coordinates (meets accuracy)
• Project plans (did not meet)
• Existing cartography (did not meet)

29
Data Collection Methods Tested by Consultants for
Iowa DOT

• Anchor sections
• Videolog van DMI (reference dataset)
• Videolog van DGPS (did not meet)
• Heads-up digitizing of 24-inch orthophotos for
  distances (did not meet)
• Heads-up digitizing of 6-inch orthophotos for
  distances (meets accuracy)
• Project plans (did not meet)
• Existing cartography (did not meet)

30
ISU Research Team Results

• All datasets but 1-meter meet anchor point
  accuracy requirements
• In progress -- distance measurements
• DMI (reference)
• Roadware van
• 2-inch photos
• 6-inch orthophotos
• 24-inch orthophotos
• 1-meter orthophotos

• All datasets meet positional accuracy
  requirements for business data

31
Creation of Datum Using Aerial Photos
Heads-up digitized centerline for datum
Cartography centerline
6-inch resolution aerial photos
32
Passing Zones on Rural 2-lane Roadways

• Provide guidance to drivers as to whether the
  geometric layout of the roadway allows sufficient
  sight distance for a following vehicle to pass a
  slower moving one
• Are identified by pavement marking
• Inventory of passing zones useful for
• Safety analysis
• Design
• Evaluation of deficiencies
• Capacity studies
• Roadway maintenance and rehabilitation

33
Identification of Passing Zones
Identification of changes in pavement marking for
passing zones

• Begin and end point represented as points
• Attribute tables created
• Point features snapped to street centerline
• Distance from reference location calculated

? Point feature to delineate changes in pavement
marking
Street database centerline
34
Using Point Features and Attributes to Linearly
Reference Passing Zones