atlas

Airborne Traffic Flow Data

and Traffic Management

Mark Hickman and Pitu Mirchandani

University of Arizona, ATLAS Center

Greenshields 75 Symposium

July 8-10, 2008

atlas

Outline

• Basics of airborne traffic data

• Macroscopic traffic flow characteristics

– Historical perspectives

– Current applications

• Microscopic traffic flow characteristics

– Historical perspectives

– Current research activities

• A look ahead

atlas

Basics of Airborne Traffic Data

• Use video and camera images, and in some cases

automated methods, to measure traffic variables

• Develop analysis techniques and measures of flow

based on spatial characteristics of traffic patterns

• Goal: Improve efficiency of transportation system by

integrating airborne data with ground-collected data

atlas

Basics of Airborne Traffic Data

Major Characteristic: We are no longer restricted to point detection,

and we can use a spatial detection paradigm

Media: Video and camera images

MEDIA

Sensor

Traffic

flows

Research effort: Use video and camera images, image processing and

algorithms to measure traffic variables

atlas

Macroscopic Flow: Historical Perspective

A. N. Johnson (1928), “Maryland Aerial Survey of

Highway Traffic Between Baltimore and Washington,”

Proceedings of the Highway Research Board.

Washington, D.C., Vol. 8, pp. 106–115.

Johnson was Dean of Engineering at the University of

Maryland and was investigating the traffic impacts of

widening of the two-lane highway between Baltimore

and Washington

atlas

Johnson’s

Aerial

Photography

Source: Johnson (1928)

atlas

Historical Perspective: Johnson

Source: Johnson (1928)

atlas

Historical Perspective: Johnson

Source: Johnson (1928)

atlas

Historical Perspective: Greenshields

“[H]igh altitude haze, shadows of buildings, trees, and the

movement of the blimp are difficulties which, it is believed, are

overshadowed by the complete and accurate record of all that

happens within the area studied” (p. 291)

“It is felt that several hours of such observations will reveal more

than days of less complete data. From this standpoint it could well

be that aerial photographs will prove comparatively cheap” (p. 297).

B. D. Greenshields (1947), “The Potential Use of Aerial Photographs in Traffic

Analysis,” Proceedings of the Highway Research Board. Washington, D.C., Vol. 27,

pp. 291–297.

Historical Perspective: Wagner and May (1963)

Traffic density contour maps from helicopter observations

atlas

atlas

Historical Perspective

Research to identify macroscopic characteristics:

• Forbes and Reiss (1952): traffic flow from video

• Jordan (1963): freeway speed, flow, density

• Rice (1963): urban traffic congestion due to access, incidents

• Cyra (1971): freeway volumes and speeds

• Makigami et al. (1985): freeway speed, density, bottlenecks

• Angel et al. (2002 onward): travel times, intersection delays

• Chandnani and Mirchandani (2002): speeds

• Agrawal and Hickman (2004): queue lengths

• Coifman et al. (2004, 2006): origin-destination flows, intersection

delay, parking lot utilization

• Etc.

atlas

Technology

UAS’s (or UAV’s): Coifman et al. (2004, 2006)

Source: Coifman et al. (2006)

atlas

Microscopic Flow Characteristics

from Airborne Data

Data

Collection

Video Image

Processing

Trajectory

Processing

Application

Post-processing

Raw Video Vehicle in Image

Vehicle Position

and Time

Registration

Vehicle detection

Vehicle tracking

Scaling Road mask

Source: Hickman and Mirchandani (2006)

Correspondence Projection

atlas

Microscopic Flow: Historical Perspective

Manual data reduction

• Treiterer et al. (1966 onward)

• Smith and Roskin (1985)

Automated data reduction

• University of Arizona / ATLAS and the Ohio State University

• Technical University of Delft

• DLR

Microscopic Flow: Treiterer

Source: Treiterer and Taylor (1966)

atlas

University of Arizona / ATLAS

VEHICLE DETECTION

CONNECTED COMPONENT

LABELING

TRACKING

DISPLAY

IMAGE REGISTRATION

atlas

Vehicle Tracking in Imagery

• Combination of short-term and long-term tracking of

connected components (“blobs”) in registered

imagery

• Short term: Components are investigated to detect

moving vehicles and screen out false positives

• Long term: Once identified, a blob should not be

“lost” in the image sequence; a location predictor is

also used

atlas

IMAGE 1

IMAGE 2

atlas

Registered Video with Tracking

Aerial Video

I-10 at Elliott in Phoenix, AZ

atlas

University of Arizona / Ohio State University

Current research:

• Investigating of truck movements at border crossings

and at work zones

• Using ground counts, selected vehicle identification

(GPS or license plates), and airborne observations

• Investigating of travel times, delays, queuing in order

to develop strategies for improved operations

Frame 1300 (0 sec)Mariposa (Arizona – Sonora) Border

atlas

I-10 Work Zone, Tucson AZ

atlas

DLR: ANTAR and

TrafficFinder

Source: Ruhe et al. (2007)

atlas

DLR

199965

199970

199975

199980

199985

199990

199995

200000

200005

200010

200015

200020

200025

200030

200035

200040

200045

200050

200055

0 250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000 3250 3500 3750

Distance along Roadway (m)

Tim

e (

Se

c)

Source: Hipp (2006)

atlas

DLR

Congested flow

Synchronized flow

Impeded free flow

Free flow

Source: Ruhe et al. (2007)

atlas

A Look Ahead

• Airborne imagery has a long history in traffic analysis

– Spatial paradigm for data collection

– Allows collection of a wide variety of performance

measures

• Macroscopic flow characteristics are regularly

collected this way today

• Tools for individual vehicle tracking and analysis are

available

atlas

A Look Ahead

• Mechanisms for data collection and reduction exist

– Macroscopic flow characteristics

– Microscopic flow characteristics

• Applications

– Congestion studies

– Real-time traffic management

– Understanding of traffic flow

– Calibration of traffic simulation models