Documente Academic
Documente Profesional
Documente Cultură
available at www.sciencedirect.com
a r t i c l e
i n f o
a b s t r a c t
Article history:
Laser scanners are increasingly used in automation and robotic applications as a sensing
device for navigation and safety. They have agricultural applications in measuring plant
growth rate, tree volume, tree count, 3D imaging, and pattern recognition. Laser scanners
10 July 2007
are commercially available, but there is very little published information on characteristics
and performance of these laser scanners. This study compared two laser scanners, the Sick
LMS200 and the Hokuyo URG-04LX, for measurement drift over time, the effect of material
Keywords:
Laser scanner
and color on measurement accuracy, and the ability to map different surface patterns.
Measurement drift over time was studied by determining the distance between the laser
Measurement drift
scanner sensor and a stationary object at different xed distances and angles. Distance
Measurement accuracy
measurements over time uctuated with a peak-to-peak value of 1020 mm. The settling
time, which is the time required for the averaged distance data to reach a stable level,
increased when measurement distance increased but for a given distance, the settling time
remained constant for different angles. At the measurement angle of 90 , the settling times
for the LMS200 and the URG-04LX for 50% of the maximum scanner measurement distances
were 53 min and 70 min, respectively. Therefore, to obtain accurate distance measurements,
the laser scanners should be warmed up for the duration of the settling time before recording
measurement data.
The measured distance for soft material objects, such as a styrofoam plate and a sheet
of dry sponge, was longer than the actual distance. For shiny objects, such as orange tree
leaves, transparency lm, and a stainless steel plate, the measurement distance was shorter
than actual distance. At the measurement angle of 90 , the difference between the longest
and shortest measured distance (90% of the maximum scanner measurement distance) was
21.3 mm for the LMS200 and 29.7 mm for the URG-04LX. At the measurement angle of 45 ,
this difference increased to 73.2 mm for the LMS200; the URG-04LX was not able to detect
any objects at 45 .
The surface shapes of a cylindrical pipe, a folded cardboard plate with a square-shaped
valley, and a folded cardboard plate with a V-shaped valley were well-depicted by the laser
scanner. For the object with a V-shaped valley with a true depth of 6.1 cm, the averaged
depths measured by the LMS200 and URG-04LX were 6.8 cm and 3.6 cm, respectively. The
larger discrepancy in the URG-04LX depth measurement may be caused by the relatively
lower angular resolution of the URG scanner, compared to that of the LMS scanner.
2007 Elsevier B.V. All rights reserved.
Corresponding author. Tel.: +1 863 956 1151x1228; fax: +1 863 956 4631.
E-mail address: ehsani@u.edu (R. Ehsani).
0168-1699/$ see front matter 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.compag.2007.08.007
c o m p u t e r s a n d e l e c t r o n i c s i n a g r i c u l t u r e 6 0 ( 2 0 0 8 ) 250262
1.
Introduction
Jimenez
et al. (1999) built a laser scanner-based measurement
system to recognize fruits in eld tree conditions, considering it as a sensing device for a fruit-harvesting robot. The
scanner provided the distance to an object and the attenuation of laser signal which occurred in the round-trip travel
to the object. The information obtained was merged to recognize the fruit and nd the nal fruit position. Hebert (2000)
compared the characteristics of several range sensing technologies used in robotics. The measurement range of a laser
scanner using the TOF principle was relatively long, compared to other technologies. The scanner provided relatively
stable, accurate measurements under hostile environmental
conditions such as fog, dust, or smoke. Monta et al. (2004)
built a three-dimensional sensing system, composed of a laser
scanner and a scanner table moving vertically, for an agricultural robot. The sensing system could detect objects such as
tree trunks, branches, and leaves in a vineyard, and calculate
the diameter of the tree trunk and the distance between the
251
2.
Objectives
The overall goal of this study was to analyze and compare the
characteristics of two commercially available laser scanners.
The specic objectives were: (i) to test distance-measurement
drifts over time at different measurement distances and
angles, (ii) to examine the effect of different materials and
colors of target objects on distance measurements, and (iii)
to measure accuracy in mapping the surface patterns of the
objects of different shapes.
3.
252
c o m p u t e r s a n d e l e c t r o n i c s i n a g r i c u l t u r e 6 0 ( 2 0 0 8 ) 250262
3.1.
Laser scanners
3.1.1.
LMS200
3.1.2.
URG-04LX
24 15%
830
4.5
185 (W) 156 (L) 210 (H)
URG-04LX
4
240
0.36
100
1
10: up to 1 m distance; 1% of distance:
14 m distance
RS232 (19.2 kbit/s, 57.6 kbit/s, 115.2 kbit/s,
500 kbit/s, and 750 kbit/s), USB 2.0
(12 Mbit/s)
5 5%
500
0.16
50 (W) 50 (L) 70 (H)
c o m p u t e r s a n d e l e c t r o n i c s i n a g r i c u l t u r e 6 0 ( 2 0 0 8 ) 250262
253
with a baud rate of 19.2750 kbit/s and a USB port with a baud
rate of 12 Mbit/s. In this study, the scanner communicated with
the computer via a USB port.
3.2.
3.2.2.
3.2.1.
Experimental methods
Drift test
For examining the effect of different materials and colors of objects on distance measurement, 11 target objects
(30 cm 30 cm) were used. The materials included ve sheets
of colored paper with a 0.5 mm thickness (white, blue, yellow,
red, and black), a 20-mm thick laminated wood plate, a 2-mm
thick stainless steel plate, a 26-mm thick styrofoam plate, a 25mm thick sheet of sponge, a 4-mil thick sheet of transparency
lm, and a sheet made of orange tree leaves that were afxed
on a transparency lm using double-sided tape without any
free space.
The three intended distances from the scanner to the
object used in the experiment for drift measurement were
also used in this experiment. The target objects were placed at
angles () of 45 and 90 with same distance (D) to the scanner
(Fig. 2b). The objects were turned towards the front so that the
effect of an incidence angle from the laser beam to the surface
of the object on distance measurement could also be studied.
The distance data to the 11 objects were collected rst at 90
for 90% of the maximum scanner measurement distances. The
experimental order of the objects was randomly determined,
and then the test was conducted again at 45 with the same
distance in the new random experimental order of the objects.
These tests were repeated at 10% and 50% of the maximum
scanner measurement distances.
In order to analyze the data, a multiple comparison analysis
was conducted using the multicompare function in MATLABs statistics toolbox (The MathWorks Inc., Natick, MA, Ver
5.1). The multicompare function follows Tukeys procedure,
which is based on the Studentized range distribution.
3.2.3.
Fig. 2 Schematic of the experimental setup (top view) (a) for drift measurement and (b) for testing objects of different
materials and colors.
254
c o m p u t e r s a n d e l e c t r o n i c s i n a g r i c u l t u r e 6 0 ( 2 0 0 8 ) 250262
Fig. 3 Objects for generating surface patterns: (a) cylindrical pipe, (b) folded cardboard plate with square-shaped valleys,
and (c) folded cardboard plate with V-shaped valleys. Distances in cm.
(1)
(2)
4.
4.1.
Measurement drift
c o m p u t e r s a n d e l e c t r o n i c s i n a g r i c u l t u r e 6 0 ( 2 0 0 8 ) 250262
255
Fig. 6 Distance measurements by the LMS200 at an intended distance of 4.0 m for the angles: (a) 45 , (b) 90 , and (c) 135 .
256
c o m p u t e r s a n d e l e c t r o n i c s i n a g r i c u l t u r e 6 0 ( 2 0 0 8 ) 250262
Fig. 7 Distance measurements by the LMS200 at 135 for the intended distances: (a) 7.2 m, (b) 4.0 m, and (c) 0.8 m.
object were setup before the test. The settling times at 0.8 m,
4.0 m, and 7.2 m distances were 32 min, 53 min, and 137 min,
respectively, showing the settling time increases when the
measurement distance increases. The differences between
the averaged distance at 2 min of run time and that at the
settling times were 2.4 mm, 5.7 mm, and 14.1 mm at 0.8 m,
4.0 m, and 7.2 m, respectively. This shows a measurement
error caused by insufcient warm-up time of the laser scanner
increases when measurement distance increases.
From this experiment, it was shown that in order to measure the distance to an object accurately, the LMS200 needs
to warm up for some time before measurement. The required
warm-up time differs depending on measurement distance
and the data measured should be averaged over a specic time
interval.
Distance measurements by the URG-04LX at a single angle
of 135 for three intended distances of 0.4 m, 2.0 m, and 3.6 m
are shown in Fig. 8. Like the distance measurements by the
LMS200 (Fig. 7), the distance data shown in Fig. 8 also uctuated with a peak-to-peak value of about 1015 mm. The
amplitude of the uctuation tended to be larger when the
measurement distance increased, but the period of the uctuation was not recognizable. The distance data, averaged
every 20 min, presented a different pattern with that of the
averaged distance data measured by the LMS200 (Fig. 7). The
averaged distances increased in the beginning of the scanners operation at the intended distances of 0.4 m and 2.0 m,
which had settling times of 50 min and 70 min, respectively.
At an intended distance of 3.6 m, the averaged distance began
to decrease, reached a bottom limit at 70 min, increased
until 111 min, and then stabilized. The averaged distances
at the settling times were quite different with the intended
distances. This might also have been caused by inaccurate
distance setup between the scanner and the object before
the test. The settling time was larger when the measurement
distance increased. The differences between the averaged distance at 2 min of run time and that at the settling times were
9.3 mm, 12.3 mm, and 4.1 mm at intended distances of 0.4 m,
2.0 m, and 3.6 m, respectively. Thus, like the LMS200, the URG04LX should also be warmed up for the settling time before the
test to minimize measurement error, and the distance data
should be averaged. In the test with the URG-04LX at a single
c o m p u t e r s a n d e l e c t r o n i c s i n a g r i c u l t u r e 6 0 ( 2 0 0 8 ) 250262
257
Fig. 8 Distance measurements by the URG-04LX at 135 for the intended distances: (a) 3.6 m, (b) 2.0 m, and (c) 0.4 m.
4.2.
Effect of different materials and colors of objects
on distance measurements
The previous experimental results demonstrated that some
warm-up settling time is required for the laser scanners to
provide stable distance measurements (Figs. 68). The highest
settling time for the LMS at a distance of 7.2 m was 137 min.
The longest period of the uctuating distance data was about
1520 min in the stable region. Thus, to avoid error in the
distance measurement by insufcient warm-up time of the
scanners and a short data-sampling period, the tests were
started after running the scanners for 3 h without data collection. The scanners recorded 1000 readings on each object,
which roughly corresponded to one period of the uctuating
distance data in the stable region.
Tables 2 and 3 show the mean and standard deviation
of distance measurements to each object by the LMS200 at
intended distances of 0.8 m and 7.2 m, respectively, and the
258
c o m p u t e r s a n d e l e c t r o n i c s i n a g r i c u l t u r e 6 0 ( 2 0 0 8 ) 250262
Table 2 Mean, standard deviation, and multiple comparison analysis of distance measurements to the different objects
by the LMS200 at an intended distance of 0.8 m
45
Material
Mean (mm)
Tree leaves
Blue paper
Black paper
Yellow paper
Red paper
Wood plate
White paper
Steel plate
Styrofoam
Sponge
Transparency
788.4
789.4
790.2
790.4
791.8
792.1
793.8
799.9
802.3
811.1
S.D. (mm)
3.8
3.8
1.7
1.6
5.0
2.5
1.7
3.9
1.6
2.5
90
Multiple comparison*
a
b
c
c
d
d
e
f
g
h
Mean (mm)
S.D. (mm)
788.2
787.5
786.3
785.3
787.2
788.2
788.3
772.1
796.7
802.5
794.4
3.3
4.7
5.0
2.0
4.8
3.5
1.9
1.6
1.9
2.0
9.8
Multiple comparison*
d
cd
c
b
cd
d
d
a
f
g
e
ah represent the alphabet indexes of the multiple comparison analyses. When the objects have the same index, there is no signicant
difference. The condence level of the multiple comparison analyses was 95%.
The mean and standard deviation of distance measurements to the different objects by the URG-04LX at intended
distances of 0.4 m and 3.6 m and the results of multiple comparison analysis are shown in Tables 4 and 5. As observed in
Tables 2 and 3, the shortest distance measurements were also
found with shiny objects and the longest distance measurements were also found with soft objects. The URG-04LX could
not detect transparency lm at 45 for an intended distance
of 0.4 m. The capability of the URG-04LX was not enough to
detect any objects at 45 for an intended distance of 3.6 m.
At an intended distance of 0.4 m (Table 4), the differences
between the longest distance observed at the soft objects and
the shortest distance at the shiny objects at 45 and 90 were
55.4 mm and 21.6 mm, respectively. At an intended distance of
3.6 m (Table 5), the differences at 90 were 29.7 mm. The measurement angles at an intended distance of 0.4 m might not
inuence on the results of the multiple comparison analysis
for the different colored papers. However, at a measurement
angle of 90 , the result of multiple comparison analysis for the
colored papers at an intended distance of 0.4 m might be more
sensitive than that at an intended distance of 3.6 m.
Table 3 Mean, standard deviation, and multiple comparison analysis of distance measurements to the different objects
by the LMS200 at an intended distance of 7.2 m
45
Material
Tree leaves
Steel plate
Wood plate
Blue paper
Black paper
Yellow paper
Red paper
White paper
Styrofoam
Sponge
Transparency
Mean (mm)
S.D. (mm)
7168.6
7169.3
7173.1
7175.1
7175.2
7176.3
7178.6
7180.2
7185.2
7241.8
2.7
4.8
4.5
4.0
4.9
3.8
4.5
2.9
6.7
5.3
90
Multiple comparison*
a
a
b
c
c
d
e
f
g
h
Mean (mm)
S.D. (mm)
7188.0
7178.2
7192.0
7189.4
7187.9
7187.4
7188.4
7188.8
7197.7
7197.4
7176.1
5.0
4.8
2.1
1.5
2.4
2.4
2.6
2.4
3.3
2.3
3.0
Multiple comparison*
cd
b
f
e
cd
c
cde
de
g
g
a
ah represent the alphabet indexes of the multiple comparison analyses. When the objects have the same index, there is no signicant
difference. The condence level of the multiple comparison analyses was 95%.
259
c o m p u t e r s a n d e l e c t r o n i c s i n a g r i c u l t u r e 6 0 ( 2 0 0 8 ) 250262
Table 4 Mean, standard deviation, and multiple comparison analysis of distance measurements to the different objects
by the URG-04LX at an intended distance of 0.4 m
45
Material
Mean (mm)
Steel plate
Black paper
Tree leaves
Wood plate
Blue paper
Yellow paper
Red paper
White paper
Sponge
Styrofoam
Transparency
360.2
397.9
398.8
399.0
399.4
400.5
402.7
405.7
412.6
415.6
S.D. (mm)
90
Multiple comparison*
4.1
2.2
2.2
2.2
2.2
2.2
2.2
2.4
2.0
2.0
S.D. (mm)
Multiple comparison*
442.6
424.0
424.8
421.9
424.3
427.2
422.9
425.0
428.7
437.5
415.9
2.5
1.8
1.7
1.8
2.0
1.8
1.8
1.6
1.7
1.8
2.8
i
d
e
b
d
f
c
e
g
h
a
a
b
c
c
c
d
e
f
g
h
ah represent the alphabet indexes of the multiple comparison analyses. When the objects have the same index, there is no signicant
difference. The condence level of the multiple comparison analyses was 95%.
4.3.
Mean (mm)
Steel plate
Blue paper
Red paper
Yellow paper
Black paper
White paper
Tree leaves
Wood plate
Styrofoam
Transparency
Sponge
3643.3
3647.6
3648.3
3648.6
3649.1
3649.5
3649.9
3652.3
3658.1
3666.6
3673.0
S.D. (mm)
3.3
2.9
2.9
3.0
2.9
2.9
3.4
2.9
2.8
2.8
2.9
90
Material
Mean (mm)
Multiple
comparison*
a
b
c
cd
cde
cde
e
f
g
h
i
ai represent the alphabet indexes of the multiple comparison analyses. When the objects have the same index, there is
no signicant difference. The condence level of the multiple
comparison analyses was 95%.
Y = 6.70X 3.52
(R2 = 0.98)
(3)
Y = 3.86X 1.63
(R2 = 0.99)
(4)
Table 6 Sizes of the laser beam spot, approximated with a rectangle, for the LMS200 and URG-04LX at three intended
distances
Distance (m)
Rectangle
LMS200
URG-04LX
45
90
45
90
Width (cm)
Height (cm)
2.0
1.8
1.2
1.8
1.5
1.0
1.2
0.9
Width (cm)
Height (cm)
5.2
3.8
3.2
3.9
4.0
0.8
3.1
0.9
Width (cm)
Height (cm)
8.3
5.6
4.8
5.6
4.8
1.2
260
c o m p u t e r s a n d e l e c t r o n i c s i n a g r i c u l t u r e 6 0 ( 2 0 0 8 ) 250262
(R2 = 0.98)
(5)
Y = 1.06X + 1.08
(R2 = 1.00)
(6)
4.4.
Fig. 9 shows the surface pattern of a cylindrical pipe generated by the laser scanners. The pattern provided a sketch
of the surface shape of the pipe, but the resolution was not
enough to depict the pipe surface in detail. In the surface pattern generated by the LMS200, the radii of the pipe obtained
by calculating the difference between the highest distance
at the edge and the shortest distance in the middle, and
obtained using the cosine law (Eq. (2)) were 3.5 cm and 3.7 cm,
respectively. The radii were 1.4 cm and 2.8 cm, respectively,
in the surface pattern generated by the URG-04LX. The radii
obtained in the pattern by the LMS were close to the true
radius of 3.5 cm, while the radii obtained in the pattern by the
URG were underestimated. The URG scanner has the lower
angular resolution of 0.36 , compared to the LMS angular resolution of 0.25 . Thus, the URG scanner might miss the rightand left-most edges of the pipe, and this may explain the
reason why the measured radii are smaller than the true
radius.
The surface pattern of an object with at hills and squareshaped valleys generated by the laser scanners are shown in
Fig. 10. The pattern clearly reconstructed the hills and valleys.
The numbers of hills and valleys in the pattern agreed with
those in the actual object. In the pattern by the LMS200, the
widths of the hills and valleys in the middle, left-most, and
right-most were obtained using the cosine law. The widths
of the hills were 6.9 cm, 6.1 cm, and 6.1 cm, respectively; and
the widths of the valleys were 6.1 cm, 4.0 cm, and 4.4 cm,
respectively. The depths of the valleys were 7.2 cm, 7.5 cm,
and 7.8 cm, respectively. In the pattern by the URG-04LX, the
widths of the hills were 6.8 cm, 5.0 cm, and 5.6 cm, respectively; and the widths of the valleys were 6.0 cm, 4.6 cm, and
4.6 cm, respectively. The depths of the valleys were 7.2 cm,
6.6 cm, and 7.0 cm, respectively.
The measured widths of the hills in the middle were close
to the true width of 7.0 cm. The widths of the hills at the left
and right sides were smaller than the true width. Since the
laser beam generated at 90 hit the middle of the object, and
the scanners were in parallel with the object, the resolution
of the surface pattern at the left and right sides became lower
than that in the middle. This might cause the smaller widths
at the left and right sides. The measured widths of the valleys
were much smaller than the true width of 7.0 cm. When the
laser beam was projected on the area of the valley, some of
the beam was blocked by the hill before it reached the valley.
Therefore, the measured widths of the valleys became smaller.
The measured depths of the valleys were close to the true
depth of 7.0 cm.
Fig. 11 shows the surface pattern of an object with V-shaped
valleys generated by the laser scanners. The pattern reconstructed the surface shape of the object well. The numbers
of hills and valleys in the pattern agreed with those in the
actual object. The averaged depths of the valleys in the patterns by the LMS and URG were 6.8 cm and 3.6 cm, respectively.
The depth measured by the URG was much smaller than the
true depth of 6.1 cm. Since the URG scanner has a relatively
larger angular interval (0.36 ), compared to the angular interval (0.25 ) of the LMS, it might miss the crests of the hills and
the bottom limits of the valleys. This might cause to underestimate the depth of the valley. This also can be conrmed
from the fact that the crests of the hills and the bottom limits
of the valleys in the pattern by the LMS are sharper than those
in the pattern of the URG.
Fig. 9 Surface pattern of a cylindrical pipe generated by the LMS200 and the URG-04LX.
c o m p u t e r s a n d e l e c t r o n i c s i n a g r i c u l t u r e 6 0 ( 2 0 0 8 ) 250262
261
Fig. 10 Surface pattern of an object with at hills and square-shaped valleys generated by the LMS200 and the URG-04LX.
Fig. 11 Surface pattern of an object with V-shaped valleys generated by the LMS200 and the URG-04LX.
5.
Conclusions
The characteristics of two commercially available laser scanners, LMS200 and URG-04LX, were analyzed and compared
through several tests. The following conclusions can be drawn
from these tests:
262
c o m p u t e r s a n d e l e c t r o n i c s i n a g r i c u l t u r e 6 0 ( 2 0 0 8 ) 250262
by the URG was much smaller than the true depth of 6.1 cm.
This may have been caused by the relatively larger angular
interval of the URG scanner.
references
Jimenez,
A.R., Jain, A.K., Ceres, R., Pons, J.L., 1999. Automatic fruit
recognition: a survey and new results using range/attenuation
images. Pattern Recogn. 32 (10), 17191736.
Kise, M., Zhang, Q., Noguchi, N., 2005. An obstacle identication
algorithm for a laser range nder-based obstacle detector.
Trans. ASAE 48 (3), 12691278.
Monta, M., Namba, K., Kondo, N., 2004. Three-dimensional
sensing system using laser scanner. ASABE Paper No. 041158.
ASABE, St. Joseph, MI.
Subramanian, V., Burks, T.F., Arroyo, A.A., 2006. Development of
machine vision and laser radar based autonomous vehicle
guidance systems for citrus grove navigation. Comput.
Electron. Agric. 53, 130143.
Wei, J., Salyani, M., 2004. Development of a laser scanner for
measuring tree canopy characteristics: phase 1. Prototype
development. Trans. ASAE 47 (6), 21012107.
Wei, J., Salyani, M., 2005. Development of a laser scanner for
measuring tree canopy characteristics: phase 2. Foliage
density measurement. Trans. ASAE 48 (4), 15951601.