ARTICLE IN PRESS

Computers & Geosciences 33 (2007) 437–444

www.elsevier.com/locate/cageo

Short Note

RockSee: Video image measurements of physical features to aid

in highway rock cut characterization
Ahmed Youssefa, Norbert H. Maerzb,, Qinfang Xiangb
a
Geological Engineering, South Valley University, Geology Department, Sohag 82514, Egypt
b
University of Missouri-Rolla, 1006 Kings-highway, Rolla, MO 65409-0660, USA
Received 9 November 2005; received in revised form 25 July 2006; accepted 14 August 2006

Abstract

Maintaining highway rock cuts for the safety of the motoring public from the risk and consequence of falling rock is an
enormous task for State Departments of Transportation. The amount of work to even evaluate and prioritize the
remediation effort is prohibitive. To facilitate the prioritization of remediation efforts, a new rock fall hazard rating system
has been developed for Missouri highways. To make the process much more efficient, video logs are used screen the rock
cuts, parameters used in the system are measured on video images, and data is automatically transferred to a GIS system.
r 2006 Published by Elsevier Ltd.

Keywords: Rock hazard rating; Rock mass classification; Highway video logs; Computer-scaled measurements

1. Introduction quickly evaluated further and prioritized for reme-

It is incumbent on highway departments of
transportation (DOTs) to maintain highway rock
cuts for the safety of the motoring public and to
reduce the risk and consequence of falling rock on
lives and property. This is an enormous task, as
highways cover vast areas through differing geolo-
gical terrains. It is clearly unrealistic to remediate all
rock cuts, so efforts have to be prioritized.
Many jurisdictions now use a rock mass classifi-
cation system to streamline the process of scoring
each major rock cut (Maerz, 2000). The road cuts
with the worst scores in the classification can be
Corresponding author. Tel.: +1 573 341 6714;
fax: +1 573 341 4368.
E-mail addresses: amyoussef70@yahoo.com (A. Youssef),
norbert@umr.edu (N.H. Maerz), xiang@umr.edu (Q. Xiang).
diation.
A new risk/consequence classification system
called Missouri rock fall hazard rating system
(MORFH RS) has been developed for Missouri
highways (Youssef et al., 2003; Maerz et al., 2005).
A summary of MORFH RS is given in Appendix A.
Efficiencies are introduced into the system by
1. using digital highway video logs, imaged at
highway speeds for prescreening,
2. making measurements of key classification para-
meters on the video images using the RockSee
program, and
3. facilitating data entry of rating data, storing,
printing, and transferring reports to a GIS
system, via a personal computing device (pcd)
attached to a GPS unit.

0098-3004/$ - see front matter r 2006 Published by Elsevier Ltd.

doi:10.1016/j.cageo.2006.08.005
 ARTICLE IN PRESS
438 A. Youssef et al. / Computers & Geosciences 33 (2007) 437–444

which would call for making measurements on the

images and viewing the rock cut in the field.

3. Measurements on scaled video

The same images that can be used for video logging

and previewing can be used to measure some of the
parameters required for the hazard rating system.
Measurements can be made on single images without
extensive vehicle instrumentation and modifications.
Although not as accurate as manual measurements in
the field, the measurements are sufficiently accurate to
provide input data for a rock hazard rating system.

Fig. 1. Simple video setup, consisting of a camcorder mounted in 3.1. Measurement principle
dash of a vehicle.
The measurement concept is based on trigono-
2. Video preview and pre-screening of rock cuts metric relationships (Fig. 2) between the image size
and the object size using:
Video images are a highly efficient way to document image size d
the rock cuts along a highway. Images can be taken at Ratio ¼ ¼ ,
object size L
highway speeds, digitally recorded, and evaluated
back in the office by the engineer or geologist. Many where L is the distance between the camera and
DOTs already have special vehicles designed to travel measurement object and d the distance to the
down highways whilst obtaining video and making projection plane of the image.
other types of measurements. An example is Road- Fig. 3 shows the imaging layout. Variables used in
ware’s Automatic Road Analyzer (ARAN) described the calculations include:
by Maerz and McKenna (1999). This is a highly
sophisticated instrumented vehicle that collects mea- 1. vehicle direction vector,
surement data about objects, features, structures, and 2. mean camera direction vector (M),
landmarks located along highways and roadways, for 3. measured object endpoint vectors in the image
highway planning, managing, and maintenance. It with respect to the camera direction vector (e.g. L
consists of a video system with a precisely calibrated and R in Fig. 3 defining the width of the ditch),
high resolution video camera, a distance measuring 4. horizontal distance between the camera and the
instrument for spatial positioning, a gyroscopic edge of the road, and
geometrics system and an ultrasonic grade system 5. vertical distance between the camera and the
for precise measurement of vehicle attitude. plane of the road.
Alternatively, a simple inexpensive video setup
can be used to obtain similar results (Maerz et al.,
2003). Fig. 1 shows the hardware setup, which
consists of a digital camcorder mounted on the dash
of a vehicle. The digital video is recorded on mini-
DV tapes, and transformed to avi files using
commercially available software such as Adobe
Premieres.1
Trained engineers or geologists can review the
video footage at a computer workstation in their
office to identify problematic cuts and then decide
which sites warrant more detailed investigation,
1
Adobe Premiere is a registered trademark of Adobe Systems Fig. 2. Illustration of relationship between projection and real-
Incorporated. life object.
 ARTICLE IN PRESS
A. Youssef et al. / Computers & Geosciences 33 (2007) 437–444 439

If high-accuracy measurements are necessary,

these parameters woold require constant monitor-
ing, including the vechile attitude (pitch and roll).
Experience in developing the Roadware ARAN
Mobile Surveyor (Maerz and McKenna, 1999) has
shown that many of these parameters can be held as
constants in the calculation:

1. The assumptions are made that vehicle direction

vector is parallel to the edge of the road, vehicle
pitch and roll are negligible, and the vehicle offset
from the edge of the road is constant.
2. The mean camera direction vector (M) is fixed at
101 to the right of the vehicle direction vector.
3. The pavement reference vector (P) is user selected
by a mouse click on a location somewhere on the
horizontal plane of the roadway, defined by the
intersection of the plane of road and the vector P.
4. Subsequently, features can be measured on the
image, as long as they are located in the plane
that is perpendicular to the focal plane of the
camera, and that contains the coordinates
identified in (3) above.
Fig. 3. Geometric layout of measurement scheme.

3.2. Imaging setup

3.2.1. Camera setup

The camera setup consists of vertical and
horizontal alignment of the camera, and setting
the zoom factor on the lens. The vertical alignment
and zoom factor are set in tandem, to ensure that
the picture encompasses the top of typical road cuts
as well as the plane of the highway. Horizontal
alignment should be set to 101 to the right of the
direction of travel. This is best accomplished by
stretching a tape measure 30.0 m in the direction of
travel, stretching a second tape measure 5.29 m at
901 and to the right, placing a vertical object at that
point, and centering the camera on that object.

3.2.2. Scale calibration

Scale calibration can be done by recording an
image of an object of known dimension in the image
and calibrating off that. The portion of the image
that the scaling is valid in all subsequent images is
defined by a vertical plane, perpendicular to the
camera vector, and that passes through the point
defined by the painted white road edge line and the
vertical dotted line that is arbitrarily placed 1/3
of the way into the image from the left hand side
(Fig. 4). This scale remains constant for that
position in all images, but makes the assumption
Fig. 4. User interface, showing a yellow vertical line 1/3 of way
across image, which is automatically drawn. Intersection of
yellow line and white edge stripe can be used to define vertical
plane perpendicular to direction of travel, in which measurements
can be made.
that the roadway is straight between the vehicle and
the plane of measurement.
Alternatively, the road width, if constant can be
used as a scaling object. This allows measurements
to be made in a vertical plane perpendicular to the
camera vector, anywhere in the image (Fig. 5).
 ARTICLE IN PRESS
440 A. Youssef et al. / Computers & Geosciences 33 (2007) 437–444

Fig. 5. Plane of measurement concept. Using mouse to click on

both ends of any of these lines result in a scaled measurement.

Clearly the further away the scale and measurement

are taken the less accurate the measurement result.

3.3. Measurements

Measurements that can be made include slope

heights, slope length, and slope angle; ditch width,
ditch depth, and ditch volume; mass volumes; and
other linear measures. Measurements all need to be Fig. 6. Example of a report generated by RockSee.
made within a ‘‘measurement plane’’ as previously
described. The user simply selects the measurement
images encompassing the rock cut, and knowing the
type and clicks on the end points of the object to be
average vehicle speed, the length of the rock cut can
measured.
be determined. An example output is given in Fig. 6.
Slope measurements (in the measurement plane)
consist of measuring the height and slope face
3.4. Measurement validation
length (if not vertical) and using a trigonometric
relationship to calculate the slope angle. Ditch
A series of test measurements were conducted to
measurements (in the measurement plane) consist
evaluate the effectiveness of the measuring system.
of measuring the width and the depth of the ditch.
Twelve sites were measured. For each site, a video
Ditch volumes per linear foot are calculated by
image and corresponding manual measurements
using one of three models for calculating the cross-
were made. To determine the effectiveness of the
sectional area of the ditch: Rectangle, triangle, or
system the error percentage was calculated for each
terrace (trapezoid). Measurements of rock volumes,
parameter.
for instance volumes of loose rock, can be estimated
Errors, defined as the percentage difference
by measuring, on a vertical slope (in the measure-
between manual and image measurements, on
ment plane), the height of loose blocks, and the
average were found to be less than 10%. The
width of loose blocks close to the proximity of the
following is the average error for each type of
measurement plane. The depth of loose rock must
measurement.
be estimated, and with that the volume of loose rock
can be predicted. Any other linear measurements (in Ditch width 6.0%
the measurement plane) can be made at any time. Ditch depth 8.6%
This includes lane and shoulder widths, and heights Slope length 4.2%
of objects at the side of the road, such as retaining Slope angle 2.7%
walls. Also, by counting the total number of video Cliff height 3.9%
 ARTICLE IN PRESS
A. Youssef et al. / Computers & Geosciences 33 (2007) 437–444 441

Shoulder width 7.6% Because these measurements are used for input
Road width 2.7% into a classification system, the measurement
Measurements do have a high variability, with a accuracy is considered to be sufficient.
few errors above 10%, and occasional errors of up to
30–40% when miss-locating the edge or the bottom of
a ditch due to the obscuring effect of vegetation. Appendix A. MORFH RS
 ARTICLE IN PRESS
442 A. Youssef et al. / Computers & Geosciences 33 (2007) 437–444
 ARTICLE IN PRESS
A. Youssef et al. / Computers & Geosciences 33 (2007) 437–444 443
 ARTICLE IN PRESS
444 A. Youssef et al. / Computers & Geosciences 33 (2007) 437–444
References
Maerz, N.H., 2000. Highway rock cut stability assessment in rock
masses not conducive to stability calculations. In: Proceedings
of the 51st Annual Highway Geology Symposium, Seattle,
Washington, August 29–September 1, pp. 249–259.
Maerz, N.H., McKenna, S., 1999. Mobile highway inventory and
measurement system. Transportation Research Board, Wa-
shington, DC, Transportation Research Record No. 1690, pp.
135–142.
Maerz, N.H., Youssef, A., Xiang, Q., 2003. Digital imaging for
screening and making measurements of features on highway
rock cuts. In: Proceedings 82nd Annual Meeting Transporta-
tion Research Board, Washington, DC, January 12–16, 20pp.
Maerz, N.H., Youssef, A., Fennessey, T.W., 2005. New risk-
consequence rock fall hazard rating system for Missouri
highways using digital image analysis. Environment and
Engineering Geoscience v.xi (3), 229–250.
Youssef, A., Maerz, N.H., Fritz, M.A., 2003. A risk-consequence
rockfall hazard rating system for Missouri highways. In:
Proceedings 54th Highway Geology Symposium, September
24–26, Burlington, Vermont, pp. 175–195.