Advances in Military Technology

AiMT Vol. 7, No. 2, December 2012

Practical Experience with Distance Measurement

Based on Single Visual Camera
A. de La Bourdonnaye1, R. Doskočil2*, V. Křivánek2 and A. Štefek2
1
École de l’Air, BA 701 – Salon de Provence, France
2
University of Defence, Brno, Czech Republic

The manuscript was received on 19 September 2012 and was accepted after revision for
publication on 4 October 2012.

Abstract:
The paper deals with the problem of distance measurement of stationary target by single
visual camera in laboratory conditions. The paper describes the practical experiments
performed in order to estimate the distance of a target using visual odometry. For the
estimation of the distances, we used a camera and a target. For these estimations,
photos were taken to obtain the coordinates of reference points on the target surface.
The coordinates obtained were used in optical equations to estimate distances of the
target in the space. The paper explains the procedures made for the experiment
preparation, describes the experiments and finally analyses the results acquired. These
practical experiments establish a basis for the implementation of the position algorithm
into navigation subsystem of swarm robots.

Keywords:
Single visual camera, distance measurement, error analysis, image analysis, optical
properties.

1. Introduction
The reason for using cameras to measure the distance from the target is a creation of
a measuring subsystem for autonomous robot which will make odometry more
accurate. This functionality will be very helpful especially in robot cooperation. To
introduce this intention, a theory has to be defined and tested in applications.
This article follows the time-line described in the paper and published at the
IEEE Conference 15 th Mechatronika 2012 [1]. In the paper’s conclusion, the general
experience and conclusions of the experiment are mentioned to enhance experiment
results.

*
Corresponding author: Department of Air Defence Systems, Kounicova 65, 662 10 Brno,
Czech Republic, Telephone: +420 973 44 36 83, E-mail: radek.doskocil@unob.cz
50 A. de La Bourdonnaye, R. Doskočil, V. Křivánek and A. Štefek

The object of all experiments described in the article was to obtain the highest
possible accuracy.
The practical experiments and the first analysis were carried out by cadet De La
Bourdonnaye from the French Air Force Academy in Salon de Provence in the context
of the Project “Learning Through Research”. It is a French program for second year
students of the Master Course. Deeper analyses of the results were carried out by the
research team of the Department of Air Defence Systems.

2. Experiment Preparation

2.1. Place of Experiment

The first task was to prepare the conditions for all experiments, followed by an
assignment of the laboratory, suitable for the experiment purposes. The laboratory
conditions simulated the conditions typical for the rooms in common buildings.
Thirty three points were measured by a theodolite situated on the floor. It is a
movable, high accuracy telescope for measuring angles in both horizontal and vertical
planes. The guaranteed precision of this geographic method is 1 mm. The obtained grid
of the points contains three main lines. The distances between the lines are 2 000 mm.
The points on the lines are in minimal distance of 500 mm between two points. The
distance between the first and the last point is 6500 mm.

2.2. Model of target

The model of target is a flat board situated on the floor, equipped with consoles with
screws that enable its perpendicular positioning. The external dimension of the target
is 750 mm x 750 mm. Three reference points – 1, 2 and 3, see Fig. 1 – on the model of
target were used for the measurements. The chosen distance (500 mm) used as a
reference one, is the distance between the red points 1 and 2. The green point (point 3)
in the middle of this distance equals the height of the camera’s optical axis (269 mm
from the floor level). The vertical line in the centre of the model of target is positioned
on the points on the floor so that the model of target’s accurate position can be set.

Fig. 1 Reference points and size (in millimetres) of the model of target
 Practical Experience with Distance Measurement Based on Single 51
Visual Camera

2.3. Static support of the camera

The issue to be solved is how to keep an optical axis of the camera in the same
position during all measurements and, in addition, perpendicular to the laboratory
floor. An experimental device was invented and made by the research team. The
camera was joined with the upper plane of the device with a screw. The lower plane
was equipped with three reticles positioned as far as possible from the camera centre
of gravity of the lower plane. The paper with the millimetre resolution grid was fixed
to the laboratory floor to keep the horizontal camera orientation.
To fix the vertical axis, another auxiliary point was created. To make sure that
the vertical axis would be alike for all pictures made afterwards, the point had to be in
the longest possible distance from the camera. It was in the same height as the
reference point 3 of the model of target, and it was located on the laboratory rear wall.

Fig. 2 Experimental static support

2.4. Camera settings

The camera Nikon D70 with lens AF-S Nikkor DX was used for all practical
experiments. The main technical characteristics are in Table 1.
Tab. 1 Main technical characteristics of camera Nikon D70
Focal length 18 – 70 mm Image size 3 008 × 2 000 px
Fnumber 3.5 – 29 Sensitivity ISO 320
Self-timer 2s CCD size 23.7 × 15.6 mm
Image quality JPEG normal Focus Manually set

Important parameters were set up manually while the other ones (i.e. shutter,
aperture) were used automatically. A delayed self-timer was applied because when a
finger presses the button to shoot, the camera can slightly move and this can lead to
measurement errors. With the delay, the camera is stable while the photo is taken.
Furthermore, no flash has been utilised because otherwise errors caused by brightness
could occur.
In the text Fnumber has been used. Eq. (1) explains it:
52 A. de La Bourdonnaye, R. Doskočil, V. Křivánek and A. Štefek

f'
Fnumber , (1)
D
where f’ is focal length in image plane and D diameter of the camera lens.
The equation (1) can be also replaced by equation (2). It means that Fnumber has
direct influence on relative aperture RA
1
Fnumber . (2)
RA

2.5. Data processing

The software GIMP 2.8.0 was applied to analyse the pictures. Once the photo is taken
with the camera, it is opened with GIMP in the computer. Then, the use the zoom of
the software accuracy of one pixel is applicable. Coordinates of the three reference
points in pixels are obtained from the photo. Standard graphic format jpeg was applied
here.

3. Experiment description
This part of the article describes practical experiments which were approved and, in
addition, a process of camera model was established.

3.1. Calibration Phase

In order to set the model of camera and to find applicable mathematical relations,
a calibration phase was necessary. In the previous experiments (see Fig. 3) the model
of camera based on the physical scales of the CCD sensor was used [1]. The model
(Eq. (3)) was partially applicable, though it needed certain enhancement, hence a
correction by a heuristics constant was carried out
b b' mm
const. , (3)
s f' mm
where b (b’) is a base size in object (image) plane and s is a distance of the target.
The camera was at the point C0 on the centre line and stayed there during all
experiments. Therefore, six photos were taken in the distance from one to six meters.
The centre of the model of target was in front of the camera, in its optical axis. The
model of target was perpendicular to the optical axis of camera. The focal length was
set up to 35 mm, Fnumber 29 was chosen and hyper-focal distance had to be slightly
more than 2 000 mm. The sensitivity ISO 320 guaranteed a good quality of photos.
The camera self-timer was set for two seconds to eliminate vibration generated by the
press of finger onto the capture. Other camera parameters were set automatically.
The mathematical model of the measurement is represented by the equation:
1
s 1 881.6 [m, px]. (4)
b'
This mathematical equation is valid in the range of distances from 1 to 6 meters.
Another approach to the target distance measurement is to find a relation linking
the distance s in pixels on CCD to the distance s in millimetres in the image. This
method is based on the optical scheme in Fig. 3 and the equations presented in the
technical report [2]. This is the final equation:
 Practical Experience with Distance Measurement Based on Single 53
Visual Camera

b' mm
2 . (5)
b px
The results of the calculation are in Tab. 2.
b

Model of 1 2
target

Lens

b' s
Image
in
CCD h
plane 2' 1'

f'

Fig. 3 Spatial arrangement of the experiments

Tab. 2 Table of pixel size on CCD for different distances of model of target
Tested distance l [mm] ∆2 [mm/px]
1 002 0.008 029 267
2 003 0.008 022 451
3 004 0.007 993 694
4 002 0.007 952 456
5 003 0.007 885 078
6 004 0.007 679 727

Fig. 4 shows the results of calibration phase after which all six samples were
approximated by the curve.
We can state that the second model is valid from the distance of 2 003 mm. If
measuring is performed in shorter distances, the camera is not able to give accurate
results because of images’ distortion. The average value of ∆2 taken in the distance of
minimum 2 003 mm is ∆2 = 0.007 957 1725 mm/px. In the next part both models are
tested.
54 A. de La Bourdonnaye, R. Doskočil, V. Křivánek and A. Štefek

4
s [m]

0
0 500 1000 1500 2000 2500
b' [px]

Fig. 4 Results of the calibration phase

3.2. Validation experiment – Distance measurement

The aim of this experiment is to measure the distance between the camera and the
model of target using the picture analysis. Therefore, twenty-one pictures were taken
in the distance from 1 m to 6 m, every 0.25 m.
Tab. 3 Main parameters of the first experiment
Focal length f’ [mm] 35 smin [mm] 1 000
Fnumber 29 smax [mm] 6 000
b [mm] 500 Rotation [°] 0

Fig. 5 shows the results of distance measurement. The black lines represent the
method which calculates with mathematical model. Their error according to the sign of
partial error is –0.30 %. The average error in absolute value is ±0.43 %. Contrary, the
grey lines correspond to the model based on CCD size. In the interval between 2 and 5
meters, the error is less than ±1.0 %. In other cases errors are too big.

4. Conclusion
The paper deals with the use of a CCD camera to obtain more accurate indoor
navigation data. It should be noted that the distance measurement contains many
obstacles. The presented achievements are valid only for an exact type of camera in
combination with a particular lens. Nevertheless, the conclusions are generally valid.
Another limitation is caused by the model of target which has to be in optical
axes and perpendicular to it. However, the accuracy of distance measurement is less
than ±1.0 % for significant samples. Both models are more accurate than the model
presented in [1]. In that case the average error could be 6.28 % and maximal error
17.7 %. Hence, a significant progress has been reached.
The method based on mathematical analysis is more accurate than the model
using CCD size mainly in low distances. Its disadvantage is that it creates a relative
complication for robot embedded computers.
The presented experiments have proved the theory applied, and open the
opportunity to define measuring subsystem for odometry support. Even though there
were found some limitations, this approach can be applied. The subsystem used in
 Practical Experience with Distance Measurement Based on Single 55
Visual Camera

cooperative swarm can improve the measurement accuracy. Furthermore, it is possible

to extend its functionality to measure the position in 2D or 3D space.

4,00%

Matematical model

∆2
3,00%

2,00%
Error [%]

1,00%

0,00%
1,0 1,5 2,0 2,5 3,0 3,5 4,0 4,5 5,0 5,5 6,0

-1,00%

-2,00%
s (m]

Fig. 5 Error of distance measurement

Although the paper provides some useful algorithms, there are still many
problems which have to be solved prior to the measuring swarm robot subsystem
implementation.

References
[1] DOSKOCIL, R., FISCHER, J., KRIVANEK, V. and STEFEK, A. Measurement
of Distance by Single Visual Camera at Robot Sensor Systems. In Proceedings of
15th International Conference on Mechatronics – Mechatronika 2012, Prague:
Czech Technical University, 7 p. ISBN 978-80-01-04985-3.
[2] DE LA BOURDONNAYE, A. Role of Optics in the Cooperative Motion Robot
Task [Technical report]. Brno: University of Defence, 2012, 49 p.
[3] BERGEON, Y. Calculation of The Distance Covered by a Robot Thanks to
Image Analysis With a Two-robot Team. In Proceedings of ICMT’11 –
International Conference on Military Technologies. Brno: University of Defence,
2011, p. 849-854.
[4] BORENSTEIN, J., EVERETT, H. and FENG, L. Where am I? Sensors and
Methods for Mobile Robot Positioning [Technical report]. Ann Arbor: The
University of Michigan, 1996.
[5] CHENAVIER, F. and CROWLEY, JL. Position Estimation for a Mobile Robot
Using Vision and Odometry. In Proceedings of IEEE International Conference
on Robotics and Automation 1992, vol 3, p. 2588-2593. DOI:
10.1109/ROBOT.1992.220052.
56 A. de La Bourdonnaye, R. Doskočil, V. Křivánek and A. Štefek

[6] QIANG, L., WENHUI, Z. and JILIN, L. Realization of Odometry System Using
Monocular Vision. In Proceedings of IEEE International Conference on
Computational Intelligence and Security 2006, p. 1841-1844. DOI:
10.1109/ICCIAS.2006.295383.
[7] SCARAMUZZA D. and FRAUNDORFER F. Visual Odometry, Part I and II.
IEEE Robotics & Automation Magazine, Dec. 2011, p. 80-92 and Jun. 2012,
p. 78-90.
[8] STELLA, E., LOVERGINE, FP., CAPONETTI, L. and DISTANTE, A. Mobile
Robot Navigation Using Vision and Odometry. In Proceedings of Intelligent
Vehicles ’94 Symposium, p. 417-422. DOI: 10.1109/ISIC.1995.525115.
[9] VITEK, R. and JEDLICKA, L. Effect of the Accuracy of the Target Range
Measurement on the Accuracy of Fire. Advances in Military Technology, 2010,
vol. 5, no. 2, p. 69-83.

Acknowledgement
The work presented in this paper has been supported by the Ministry of Defence of the
Czech Republic in institutional development project “Development of air forces
constituents in connection with simulation technologies and cybernetic systems”.