Vehicle Speed from Aerial Photography:

Accuracy Analysis

Prof. Ismat M. Elhassan

King Saud University

Riyadh, Saudi Arabia
E-mail: [email protected]
 Presentation
- Introduction
- Importance of Vehicle Speed
- Effects of Vehicle Speed
- Techniques of Vehicle Speed Determination
- Vehicle Speed from Aerial Photography
- Previous Studies
- Accuracy Analysis
- Conclusions
 Introduction
Definition:
•Speed is defined as the rate of motion (distance
per unit of time).

* Mathematically, speed or velocity, v is given

by:
v=d/t
Where d is distance traveled during a unit time t
* Speed of different vehicles will vary with
respect to time and space.
 Importance of Vehicle Speed
*A basic measure of traffic performance and
road safety
* Establishing speed limits,
* Determining safe speed limits at curves,
* Establishing lengths of non-passing zones,
* Location of traffic signals,
* School or hospital zone protection,
* Geometric design features: lengths of speed
change lines, sight distance evaluation.
 Effects of Vehicle Speed
* Crashes of all types:
- According to NHTSA’s 2011 Fatality
Analysis Reporting System , high vehicle
speed was a contributing factor in 29 % of all
fatal crashes in USA.
- In 2011, 32,367 people died in motor
vehicle traffic crashes in the USA.
- In 2011, an estimated 2.22 million people
were injured in motor vehicle traffic crashes.
 Car Accidents and Fatality in KSA

According to A. Al-Ghamdi (1996):

Fatality Injury Accidents Reg. Population Year

Vehicles (millions)

834 6530 7197 180185 6.5 1972

2953 18616 21597 3018811 11.062 1982
3495 27385 40076 5328505 16.328 1992
 Increase of Distance to Stop a Vehicle
High Speed increases distance needed to stop
a vehicle
 Increasing Crash Energy
* Increase in speed leads to exponential
increase of crash energy: a 50% increase in
speed causes impact energy to increase by
125%
* A pedestrian hit at 64.4 km/h has an 85-
percent chance of being killed; at 48.3 km/h
the likelihood goes down to 45 percent, while
at 32.2 km/h, the fatality rate is only 5
percent (U.K. Department of Transport).
 Increasing Fatality Risk
The fatality rate of pedestrians in crashes with passenger
cars as function of the collision speed (Rosén et al., 2011).
 Increasing Crash Frequency
Traffic Speed and Crash Frequency
Techniques for Measuring Vehicle Speed
RADAR: Shift in Frequency by Doppler Effect

Sources of Errors: Cosine error, frequency

.drift due to temp change and volt fluctuation
.Radar units must be calibrated
 LIDAR Speed Detector
LIDAR : LIght Detection And Ranging.  A police
laser emits a highly focused beam of invisible
light, in the near infrared region of light, that is
centered at 904nm of wavelength and is only
.about 56cm in diameter at 300m

Police laser-lidar calculates speed by observing

the changing amount of time it takes to "see"
reflected pulses of light over a discreet amount
.of time
 GPS
GPS devices are positional speedometers. GPS
algorithm also uses the Doppler shift in the
.pseudo range signals from the satellites

It should also be noted that the speed reading is

normalized, and is not an instant speed.

Accuracy of calculated speed, is dependent on

the satellite signal quality at the time.
 Overlapping Aerial Photographs
Overlapping aerial photos allow capturing
positions of moving vehicle at two different
instants known as time interval of aerial Photos.
Howes, William F and Miles, Robert Douglas, in
1963, published a technical paper titled:
"Aerial Photography Applied to Traffic Studies"
in report number FHWA/IN/JHRP-63/14,
Purdue University, USA.
 Speed from Digital Aerial Photography
From: Fumio Yamazaki et al, 2007:

Two consecutive aerial images of central Tokyo having

about 3.08 s time lag, taken by digital aerial camera.
The images were provided by Geographical Survey
Institute of Japan.
 Continuous Strip Aerial Photography
In 1965 McCasland reported on comparison of
two techniques of aerial photography for
application in freeway traffic operation studies:
using aerial strip photography (ASP) and aerial
time lapse photography.

ASP provided more coverage while more speed

readings can be given by time lapse
photography.
 
 Aerial Video
Pitu Mirchandani, et. al, 2002 reported on aerial video
application on traffic performance.
The estimated speeds of the matched vehicles were
calculated using the image scale and the time between
frames.
Vehicle speeds were fairly well clustered around 48-53 mph,
as one might expect along a freeway.
The lone exception was vehicle 9 which had an observed
speed of over 70 mph. This anomalous speed was due to the
erroneous matching of vehicle 9 with vehicle 10.
Such an erroneous speed could be discovered by a simple
filter.
 Satellite Imagery
Zhang and Xiong (2006) and presented a
technique to detect a moving vehicle using one
single set (Panch. and MSS sensors)of Quickbird
imagery.
The test results showed that the position, speed
and moving direction of a vehicle can be
determined at a reasonable accuracy.
The speed accuracy depends on the image
resolution, accuracy of vehicle image coordinates
and accuracy of satellite time interval between
Panchromatic and Multispectral images.
 Photogrammetric Approach
The overlapping photos method:
Assuming time lapse between the overlapping
photographs is t (seconds), image distance between
the positions of the vehicle on one of the photographs
is measured as: d (mm), flying height H and camera
focal length f, are given in meters and mms,
respectively:

vehicle speed, v = (d*H/f) / (t) m/sec

 Time lapse Overlapping photography
Two aerial photos exposed from stations O1 and
O2:
O1 O2

c1 c1 c2

C1 C2

Car position C1 moved to positionC2 during time lapse

between photos
 Vehicle speed formula
The vehicle speed v (m/s) can be calculated
from the time-distance relation:
v = D / t, or v = d (H/f)/ t,
where t is the time interval between the
consecutive photos in seconds.
To determine the speed in kilometers per
hour:
v = d (H/f)/ t * (3600/1000) or:
v = 3.6 * d (H/f)/ t
 Variance of computed speed

Given a function F(x,y) in wich both arguments

x and y contain statistical errors and are
statistically independent, the variance for the
value of F(x,y) is obtained by law of
propogation of errors:
var(F) = (∂F/∂x)2 var(x) + (∂F/∂y)2 var(y).
Stand. Error: dF = var(F(x,y))1/2.
This law is applied to determine error in car
speed:
 Standard Error
in Vehicle Speed
The function, F, is the vehicle speed, v.
The standard error in v will be given as:
dv = var(v(d, H, f, t ))1/2.
var(v) = (∂v/∂d)2 var(d) + (∂v/∂H)2 var(H) +
(∂v/∂f)2 var(f) + (∂v/∂t)2 var(t)
Where:
∂v/∂d = 3.6 (H/f) / t, ∂v/∂H = 3.6 (d/f) / t,
∂v/∂f = 3.6 * [d (H)/ t] / (f2),
∂v/∂t= [3.6 * d (H/f)] / (t)2
 Tested Data - 1
DATA 1: Vehicle speed v = 50 km/h, 100km/h, 150km/h
Photo scale: 1/5000, t = 5 sec. σd = 0.01mm
Effect of change in σt (accuracy of time interval)
Results:
σv (k/h) σv (k/h) σv (k/h) σt (sec)
V = 150 k/h V = 100 k/h V = 50 k/h
1.50 1.00 0.50 0.05
3.00 2.00 1.00 0.10
6.00 4.00 2.00 0.20
15.00 10.00 5.00 0.50
 Errors in v scale 1:5000

dt= 5 sec, Sigma d=0.01mm, v1=50k/h, v2=100;/h,

v3=150k/h, standard error in k/h
Sigma v
16
14 Sigma dt
12
0.05
10
0.1
8
6 0.2
4 0.5
2
0
v1 v2 v3

50k/h 100k/h 150k/h

 Test Data - 2
DATA 2: Vehicle speed v = 50 km/h, 100km/h,
150km/h
Photo scale: 1/10000, t = 5 sec. σd = 0.01mm
Results:
σv (k/h) σv (k/h) σv (k/h) σt (sec)
V = 150 k/h V = 100 k/h V = 50 k/h
1.50 1.00 0.50 0.05
3.00 2.00 1.00 0.10
6.00 4.00 2.00 0.20
15.00 10.00 5.00 0.50
 Test Data - 3
DATA 3: Vehicle speed v = 50 km/h, 100km/h,
150km/h
Photo scale: 1/5000, t = 5 sec. σd = 0.01mm
Effect of change in σd (accuracy of mesuring vehicle
movement on photo), Results:
σv (k/h) σv (k/h) σv (k/h) σd (mm)
V = 150 k/h Scale Scale
1/10000 1/5000
1.50 0.01 0.007 0.002
3.00 0.07 0.040 00.10
6.00 0.36 0.180 0.050
15.00 0.72 0.360 0.10
 Graphical Representation
Effect of change in sigma d and scale
 Conclusions
• From the above tests and results it can be concluded
that:
• *To get high vehicle speed accuracy we need to
have:
• Highly accurate measured time interval between
photos (intervalometer)
• Highly accurate photo measured distance moved by
vehicle.
• Large scale photography
* The advantage of using aerial photography is that the
information is available any time needed.
 END
• THANKS

• For all audience

• Have good time