See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/224245297

Low cost colour sensors for monitoring plant growth in a laboratory

Conference Paper  in  Conference Record - IEEE Instrumentation and Measurement Technology Conference · June 2011
DOI: 10.1109/IMTC.2011.5944221 · Source: IEEE Xplore

CITATIONS READS

22 4,408

4 authors, including:

Gourab Sen Gupta Donald G Bailey

Massey University Massey University
169 PUBLICATIONS   1,417 CITATIONS    227 PUBLICATIONS   2,277 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Apple breeding View project

Embedded Systems View project

All content following this page was uploaded by Donald G Bailey on 19 May 2014.

The user has requested enhancement of the downloaded file.

 Low Cost Colour Sensors for Monitoring Plant
Growth in a Laboratory
Mark Seelye, Gourab Sen Gupta, Donald Bailey John Seelye
School of Engineering and Advanced Technology (SEAT) New Zealand Institute for Plant & Food Research Limited
Massey University, Palmerston North, New Zealand Private Bag 11 600, Palmerston North, New Zealand
Email: [email protected], [email protected]

Abstract— An automated system for measuring plant leaf colour, autonomous wheeled robot has been developed to pick
as an indicator of plant health status, has been developed for kiwifruit from orchard growing vines [7]. Robotic techniques
plantlets growing in a modified micropropagation system. Using for production of seedlings have been developed, identifying a
a custom built robotic arm, sensors located on a pan and tilt need to add a machine vision system to detect irregularities in
system at the end of the arm monitor plant growth and the seed trays and to provide supplementary sowing using a 5-arm
ambient growing environment. Sensors include a compact colour robot [8].
zoom camera, RGB (red, green and blue) colour sensors, and
environmental sensors. Leaf colour sensors provide information, Automation of micropropagation for the rapid
in a non-destructive manner, on the health status of tissue by multiplication of plants has been described for the
comparing the sensor outputs to pre-determined optimum values. micropropagation of a grass species that replaces the costly and
These low cost colour sensors can be incorporated into a tedious manual process [9]. A system has also been developed
continuous automated system for monitoring leaf colour of that combines plant recognition and chemical micro-dosing
growing plants. Subtle colour changes can be an early indication using autonomous robotics [10].
of stress from less than optimum nutrient concentrations. When
combined with automated image sensing for growth analysis, and Colour as a means of assessing quality is also gaining
environmental sensing (RH, CO2 and temperature) in a popularity amongst researchers. These include evaluating
controlled environment, optimised rapid growth with minimal bakery products using colour-based machine vision [11],
human input can be achieved using a modified micropropagation monitoring tea during fermentation [12], grading specific fruits
system. In this paper we detail the calibration technique for a and vegetables [13, 14, 15] and in the health sector to
RGB sensor and compare it with a high end spectrophotometer. determine blood glucose concentrations [16]. Near infrared
(NIR) sensors are also gaining popularity as non-destructive
Keywords - plant growth monitoring, colour sensor calibration, means of assessing fruit and plant material, including use as a
RGB colour sensor measure of plant nutrient status [17] as well as testing of fruit
quality [18, 19, 20].
I. INTRODUCTION
Investigation into non-destructive methods to measure the
Robotic and automated systems are becoming increasingly health status of plants is a relatively new area of research.
common in all economic sectors. In the past decade there has Subtle leaf colour changes can be used as a measure of plant
been a push towards more automation in the horticulture health. Although limited work has been carried out in real time,
industry, and it is only now, as robots become more a recent micropropagation-based system used potato tissue
sophisticated and reliable, that we are beginning to see them images captured via a digital camera then scanned to identify
used to undertake routine, often repetitive tasks, which are the colour of selected pixels [21]. Spectral reflectance, using a
expensive to do using a highly paid labour force. With rapid range of spectral bands, has been used as a non-destructive
strides in technological advancement, more and more measure of leaf chlorophyll content in a range of species [22].
applications have become possible. These include the Alternative methods make use of spectroscopic systems using a
development of a robotic system for weed control [1], a system fixed light source to record colour reflectance of multiple
for automatic harvesting of numerous agri-science products samples [23].
such as cutting flowers grown in greenhouses [2] and
automating cucumber harvesting in greenhouses [3]. Advances This paper focuses on the use of low cost colour sensors for
in electronics have empowered engineers to build robots that monitoring leaf colour of plant tissues growing in a modified
are capable of operating in unstructured environments [4]. micropropagation system. The calibration method of these
Camera-in-hand robotic systems are becoming increasingly sensors is described and its accuracy evaluated.
popular, wherein a camera is mounted on the robot, usually at
the hand, to provide an image of the objects located in the II. OVERVIEW OF SYSTEM HARDWARE
robot’s workspace [5]. Increasingly, robots are being used to The reported system uses an autonomous robotic arm
sort, grade, pack and even pick fruits. Fruits can be identified containing RGB (red, green and blue) colour, environmental
and classified on a continuously moving conveyer belt [6]. An and proximity sensors as well as a compact colour camera.
Custom software created in Microsoft® Visual Studio (VB.net) Optoelectronic Solutions) light-to-voltage chip. When light is
allows for a completely automated operation that requires reflected, the voltage, which is proportional to the light
minimal human input. reflected, is used to determine the sample’s R, G and B colour
contents. The ColorPAL requires the sample to be illuminated
A. Robotic Arm using each of the red, green and blue LEDs, with a ‘snorkel’ to
The robotic arm, shown in Figure 1, uses five stepper shield possible interference from external light sources. This
motors that are controlled through a motor controller and requires the ColorPAL to be in direct contact with the object
micro-step driver [24]. To allow the robotic arm to move for an optimum reading without interference.
vertically, a ball screw and shaft assembly is incorporated,
converting rotational motion into vertical movement. The arm
contains a pan and tilt system at its distal end, which houses a
camera [25], colour and proximity sensors [26]. The operation
of the arm is completely automated, continually gathering
information from the sensors and capturing images for
assessment and analysis.
The arm uses bipolar, high torque stepper motors, which
provide a maximum torque of 12.3 kg/cm. and have a step
angle of 1.8o. The use of a micro step driver allows the user to
select an even finer resolution (i.e. more steps per revolution).
The motors are controlled through a motor driver board that Figure 3. Parallax TCS3200 colour sensor.
allows commands to be sent from the PC via a USB port to the
controller to control the movement. The TCS3200 Colour sensor (Figure 3) makes use of a
TAOS TCS3200 RGB light-to-frequency chip. The TCS3200
colour sensor operates by illuminating the object with two
white LEDs, while an array of photo detectors (each with a red,
green, blue and clear filter) interpret the colour being reflected
by means of a square wave output whose frequency is
proportional to the light reflected. The TSC3200 Colour sensor
has a 5.6-mm lens, which is positioned to allow an area of 3.5
mm2 to be viewed.
Figure 1. SolidWorks rendered photo of the robotic Arm.
A USB4000 spectrometer (Ocean Optics Inc., FL, USA)
was used to find the height at which the greatest intensity of
B. Colour Sensors light occurred when the RGB sensor was placed above a
Currently there are a number of colour sensors on the sample. As the two white LEDs are directed down at an angle,
market, with prices ranging from low cost light-to-frequency there is a point where the light intensity is the greatest. This
chips to sophisticated and very expensive spectrophotometers. position was 20 mm above the surface of the sample, as shown
in Figure 4.
Parallax (Parallax Inc, CA, USA) has two colour sensors
that integrate seamlessly with their Basic Stamp
microcontroller. Both the ColorPAL and TCS 3200 colour
sensors are provided with some source code, making them
amenable to integrating into our customised system.

Figure 4. Light absorbed from TCS3200 across the white LED light
spectrum when the sensor is positioned at 6 different heights.

Since the TCS3200 is mounted 20 mm above the sample,

Figure 2. Parallax ColorPAL colour sensor. and therefore not in direct contact with the sample, it was more
suited for our application than the full contact required by the
The ColorPAL sensor (Figure 2) illuminates a sample using ColorPAL sensor.
in-built red, green and blue LED light sources (one colour at a
time) and records the quantity of light reflected back from the An alternate method of determining plant leaf colour is to
object. The ColorPAL makes use of a TAOS (Texas Advanced use an image captured by a camera and through software
determine the colour of the pixels. However, because plant TABLE I. X, Y, AND Z CHROMATICITY VALUES OF RED, GREEN AND
BLUE CONVERTING XYZ TO SRGB
leaves can overlap (see Figure 5), shadows are created, leading
to false colour readings from the image. Since the TCS3200 Colour x y z
colour sensor uses its own light source to illuminate the sample
surface, it eliminates any potential shadowing. Red 0.64 0.33 0.212656
Green 0.30 0.60 0.715158
Blue 0.15 0.06 0.072186
Areas of
overlapping
showing From the x, y and z chromaticity values, the transformation
“darkness” matrix, M, is calculated (Eq. 5)

⎛ 0.721144 0.063298 0.166008 ⎞ (5)

⎜ ⎟
M ≈ ⎜ 0.303556 0.643443 0.052999 ⎟
⎜ 0.000076 0.064689 1.024294 ⎟
⎝ ⎠
To calculate the R, G and B values the inverse is taken (Eq.
6 - 7).
Figure 5. Captured image showing areas of ”darkness” caused by
overlapping which would lead to false colour readings ⎛R⎞ ⎛X⎞
⎜ ⎟ −1 ⎜ ⎟
A Konica Minolta CM-700D Spectrophotometer (Konica ⎜ G ⎟ = M × ⎜Y ⎟ (6)
Minolta Sensing Americas, Inc., NJ, USA) was used to validate ⎜B⎟ ⎜Z ⎟
⎝ ⎠ ⎝ ⎠
and calibrate the RGB sensors. For accurate measurements, the
CM-700D was calibrated by taking both white and black
⎛ 1.436603 -0.118534 -0.226698 ⎞ (7)
readings by sampling a supplied white and black object ⎜ ⎟
respectively. M −1 ≈ ⎜ -0.681279 1.618476 0.026671 ⎟
⎜ 0.042919 -0.102206 0.974614 ⎟
⎝ ⎠
The CM-700D gives colour in the XYZ colour space, as
well as L*a*b*, L*C*h, Hunter Lab, Yxy and Munsell. A
linear transformation matrix was required to transform data
from the XYZ colour space to the RGB colour space to enable
comparisons and calibrations to the Parallax sensor. The linear III. EXPERIMENTAL RESULTS AND DISCUSSION
transformation equation to be solved [27] is: In order to validate the TCS3200 colour sensor, it was
necessary to calibrate and test it using the CM-700D.
This involved taking 200 RGB readings with the TCS3200
⎛X⎞ ⎛R⎞ using different coloured samples and averaging them. The
⎜ ⎟ ⎜ ⎟ same samples were measured, each 20 times, with the CM-
⎜Y ⎟ = M × ⎜ G ⎟ (1)
700D and again averaged. These tests were all completed in a
⎜Z ⎟ ⎜B⎟
⎝ ⎠ ⎝ ⎠ constant temperature dark room. As the CM-700D uses the
XYZ colour space, a linear transformation matrix was required
X to convert the XYZ values to an RGB colour space (Eq. 6).
x=
X +Y + Z (2) The TCS3200 was firstly calibrated through software by
modifying the integration time, to allow the white object (used
Y to calibrate the CMD-700) to have a RGB value as close as
y=
X +Y + Z (3) possible to 255,255,255 followed by scaling each of the RGB
values, to ensure the white reading was that of the CMD-700.
Z
z= In order to calculate a calibration factor the following
X +Y + Z (4)
equation was used: RN' = RN γ (8)
'
Where: RN = CM-700D (desired RGB value)
Equations (2 – 4) combined with the standard 1931 xy
chromaticity diagram provided the foundation for the linear RN = TCS3200 RGB (Un-calibrated sensor data)
transformation (Eq. 1). This transformation converted the XYZ
data to an sRGB colour space, with the chromaticity values of γ = Gamma (required calibration factor)
x, y and z shown in Table I being standard [28].
First the TCS3200 sensor data were scaled to ensure all
values are offset, thus ensureing that the white reading is that of
the CMD-700 for each of R, G and B (Eq. 9)
 R'N G' B' (9)
RN = R × , G N = G × N , BN = B × N
Rmax Gmax Bmax

where Rmax , Gmax , Bmax represent the maximum R, G and

B value of a white object from the TCS3200.
The calibration factors (γ) for each colour were calculated
using normalized data. (Eq. 10)

log( RN' / 255) log(GN' / 255) log( BN' / 255) (10)

γR = ,γ G = ,γ B =
log( RN / 255) log(GN / 255) log( BN / 255)

For each colour sample measured, the calibration factor

was calculated and averaged using a geometric mean (as
opposed to the more general arithmetic mean function [29]),
thus providing the γ factor for R, G and B individually. The
(desired) calibrated values were then obtained using equation
11.

RN' (calibrated ) = ( RN / 255)γ × 255 (11)

For a range of seven colours, measurements were taken

using the TCS3200 RGB sensor and the CM-700D
Spectrophotometer (Table II). The gamma calibration factors
calculated were:
(Red) γR = 0.88, (Green) γG = 0.46, (Blue) γB = 0.68
Table III summarises the average error, error percentage
and the standard deviation for un-calibrated and calibrated
RGB sensor data compared with CM-700D spectrophotometer
outputs.

TABLE II. RESULTS OBTAINED COMPARING THE TCS3200 COLOUR

SENSOR (CALIBRATED AND UN CALIBRATED) WITH THE CM-700D OVER A
RANGE OF 7 DIVERSE COLOURS

TSC3200 (un TCS3200 CM-700D

Colour
calibrated) (Calibrated) Spectrophotometer
RN GN BN RN’ GN’ BN’ R G B
Red 160 45 39 170 114 71 232 91 77
Green 94 127 59 106 184 94 14 197 76
Figure 6. TCS3200 sensor RGB readings, calibrated and un-calibrated,
Light Blue 128 179 200 139 216 216 156 217 214
compared with the CM-700D readings of: Red (A); Green (B); Blue (C).
Light Green 168 196 153 177 225 181 195 231 171 (Colour samples are as given in Table II)
Dark Blue 39 106 160 49 170 186 1 168 200
White 248 250 246 249 253 249 248 250 246
TABLE IV. AN EXAMPLE OF A GREEN COLOUR INTERPRETED BY THE CM-
Black 19 19 21 26 77 48 59 58 55 700D AND TCS3200 COLOUR SENSOR BEFORE AND AFTER A CALIBRATION
FACTOR
TABLE III. AVERAGE ERROR (0-255), PERCENTAGE ERROR AND
STANDARD DEVIATION FOR RED, GREEN AND BLUE MEASUREMENTS OF THE TCS3200 TCS3200 CM-700D
TCS3200 COLOUR SENSOR, CALIBRATED AND UN CALIBRATED, COMPARED (uncalibrated) (calibrated) Spectrophotometer
WITH CM-700D RESULTS ACROSS A RANGE OF COLOURS

TCS3200 (uncalibrated) TCS3200 (calibrated) RGB = 94,127,59 RGB = 106,184,94 RGB = 14,197,76

Colour R G B R G B
Ave
Error 40.339 40.304 21.128 38.951 8.663 10.571
Error % 15.819 15.805 8.286 15.275 3.397 4.145
σ 26.285 20.329 13.055 29.668 8.081 2.852
 As the colour sensor is intended to measure the colour of
plant leaves, there is no requirement to calibrate it across the
full range of colours. The sensor was therefore calibrated for a
range of green - yellow colours only. Fifteen RHS (Royal
Horticulture Society, London, UK) colour charts, designed for
growers to classify colours, were used, with the measured data
shown in Table V. Two hundred colour readings were again
taken (and averaged) by the TCS3200 colour sensor, followed
by 20 readings (averaged) by the CM-700D for each colour
chart.
The gamma calibration factors calculated are-
(Red) γR = 0.50, (Green) γG = 0.38, (Blue) γB = 0.59
Table VI summarises the average error, error percentage
and the standard deviation for un-calibrated and calibrated
RGB sensor data compared with CM-700D spectrophotometer
outputs for the 15 colours. There was a vast improvement
across the 3-colour components (RGB), with the average red
error improving to 5.69%, green 3.19% and blue 3.92%.

TABLE V. RESULTS OBTAINED COMPARING THE TCS3200 COLOUR

SENSOR (CALIBRATED AND UN CALIBRATED) WITH THE CM-700D OVER A
RANGE OF 15 COLOURS

Colour TCS3200 TCS3200

CM-700D
(uncalibrated) (Calibrated)
RHS Colour
ID RN GN RN RN’ GN’ BN’ R G B
Group
Blue-Green 123A 99 150 148 159 206 188 138 208 194
Green 127C 38 80 66 99 160 120 55 166 134
Green 129C 99 154 121 159 208 168 143 217 173
Green 131C 25 42 31 80 123 78 45 120 84
Green 133C 62 89 75 126 167 128 122 166 144
Green 135C 42 52 31 104 134 78 102 156 97
Green 137C 42 52 31 104 134 78 115 130 85
Green 139C 68 83 51 132 162 104 141 162 108
Green 141C 57 81 40 121 161 90 129 164 87
Green 143C 71 89 42 135 167 93 149 166 81
Yellow-Green 145C 171 171 108 209 217 157 224 224 151
Yellow-Green 147C 84 87 55 147 166 107 170 154 112 Figure 7. TCS3200 sensor RGB readings , calibrated and un calibrated,
Yellow-Green 149C 174 186 101 211 224 151 221 237 134 compared with the CM-700D readings of: Red (A); Green (B); Blue (C) for a
range of RHS green colours. (Colour samples are as given in Table V)
White 155D 255 253 228 255 254 239 255 253 228
Black 202A 17 17 18 66 86 57 62 62 61
TABLE VII. VISUAL RESULTS SHOWING THE RGB COLOUR INTERPRETED
TABLE VI. AVERAGE ERROR (0-255), PERCENTAGE ERROR AND BY THE CM-700D AND TCS3200 COLOUR SENSOR, BEFORE AND AFTER
STANDARD DEVIATION FOR RED, GREEN AND BLUE MEASUREMENTS OF THE CALIBRATION (RHS 141C)
TCS3200 COLOUR SENSOR, CALIBRATED AND UN CALIBRATED, COMPARED
WITH CM-700D RESULTS ACROSS A RANGE OF COLOURS
TCS3200 TCS3200 CM-700D
(uncalibrated) (calibrated) Spectrophotometer
TCS3200 (uncalibrated) TCS3200 (calibrated)

Colour R G B R G B RGB = 57,81,40 RGB = 121,161,90 RGB = 129,164,87

Ave Error 51.133 66.459 48.513 14.509 7.179 9.114

Error % 20.052 26.063 19.025 5.690 2.815 3.574
σ 24.628 24.030 16.998 12.303 7.473 5.374
 With the calibration factor determined, the TCS3200 was [8] Wang HonYong; Cao QiXin; Masateru, N.; Bao JianYue, “Image
able to identify the colour of selected plant material. The sensor processing and robotic techniques in plug seedling production”,
Transactions of the Chinese Society of Agricultural Machinery, Vol. 30,
has been integrated with a proximity sensor to ensure all 1999, pp. 57-62
readings are taken at a fixed height above the plant leaves. As [9] Otte, C., Schwanke, J., & Jensch, P. (1996). Automatic
the colour sensor is attached to a robotic arm, the sensor is able micropropagation of plants. Optics in Agriculture, Forestry, and
to return autonomously to the same position to record the Biological Processing, 2907, pp. 80-87.
colour, making it possible over time to identify subtle changes [10] Sogaard, H. T., & Lund, I. (2007). Application accuracy of a machine
in the colour of the plant leaves. These changes in plant vision-controlled robotic micro-dosing system. biosystems engineering,
material will allow plant nutrient concentrations to be adjusted 96(3), pp. 315-322.
in order to achieve optimum growth. [11] Abdullah, m. Z., aziz', s. A., & mohamed, a. M. D. (2000). Quality
inspection of bakery products using a color-based machine vision
system. Journal of food quality, pp. 23.
IV. CONCLUSION AND FUTURE WORK [12] Borah, S., & Bhuyan, M. (2005). A computer based system for matching
colours during the monitoring of tea fermentation International Journal
Experimental results show that the Parallax TCS3200 is a of Food Science and Technology, pp. 40.
useful low cost colour sensor, which following calibration can
[13] Omar, A. F. B., & MatJafri, M. Z. B. (2009). Optical Sensor in the
provide accurate RGB readings. It is therefore a useful Measurement of Fruits Quality: A Review on an Innovative Approach.
component for integrating into an automated monitoring International Journal of Computer and Electrical Engineering, 1(5).
system such as a robotic arm, with various other sensors, for [14] Miranda, C., Girard, T., & Lauri, P. E. (2007). Random sample estimates
the monitoring and control of plants growing in a modified of tree mean for fruit size and colour in apple. Scientia Horticulturae,
plant micropropagation system. Future work involves 112, pp. 33-41.
completing the entire robotic system with fully integrated [15] Kang, S. P., & Sabarez, H. T. (2009). Simple colour image segmentation
sensors. This will allow investigations to be carried out to of bicolour food products for quality measurement. Journal of Food
Engineering, 94, pp. 21-25.
optimise plant growth based on the information obtained from
[16] Raja, A. S., & Sankaranarayanan, K. (2006). Use of RGB Color Sensor
the colour sensors, by relating colour information to plant in Colorimeter for better clinical measurements of blood Glucose. BIME
quality. This will be further enhanced by additional growth Journal 6(1), pp. 23 - 28.
data obtained from captured plant images. The use of NIR [17] Menesatti, P., Antonucci, F., Pallottino, F., Roccuzzo, G., Allegra, M.,
sensors offers further potential for non-destructive assessments Stagno, F., et al. (2010). Estimation of plant nutritional status by Vis–
of plant tissue structure. NIR spectrophotometric analysis on orange leaves. biosystems
engineering, pp. 105.
Further test results will be presented in an extended version [18] Hu, X., He, Y., Pereira, A. G., & Gómez, A. H. (2005). Nondestructive
of this paper. It will also include investigations into a CIE RGB Determination Method of Fruit Quantity Detection Based on Vis/NIR
transformation (instead of a sRGB transformation) and the use Spectroscopy Technique. Engineering in Medicine and Biology
of a least squares regressions approach as an alternative method [19] Paz, P., S´anchez, M. T., P´erez-Mar´ın, D., Guerrerob, J. e.-E., &
to calculating the gamma calibration factor. Garrido-Varob, A. (2009). Evaluating NIR instruments for quantitative
and qualitative assessment of intact apple quality. In Wiley Interscience.
[20] Nicola¨ı, B. M., Beullens, K., Bobelyn, E., Peirs, A., Saeys, W., Theron,
REFERENCES K. I., et al. (2007). Nondestructive measurement of fruit and vegetable
[1] Slaughter, D. C., Giles, D. K., & Downey, D. (2008). Autonomous quality by means of NIR spectroscopy: A review. Postharvest Biology
robotic weed control systems: A review. Computers and Electronics in and Technology, pp. 46.
Agriculture, 61(1), pp. 63-78. [21] Yadav, S. P., Ibaraki, Y., & Gupta, S. D. (2010). Estimation of the
[2] Kawollek, M., & Rath, T. (2008). Robotic Harvest of Cut Flowers Based chlorophyll content of micropropagated potato plants using RGB based
on Image Processing by Using Gerbera jamesonii as Model Plant. image analysis. Plant Cell Tissue and Organ Culture, 100(2), pp. 183-
Proceedings of the International Symposium on High Technology for 188.
Greenhouse System Management, Vols 1 and 2(801), pp. 557-563. [22] Gitelson, A. A., Gritz, Y., & Merzlyak, M. N. (2003). Relationships
[3] van Henten, E. J., Hemming, J., van Tuijl, B. A. J., Kornet, J. G., between leaf chlorophyll content and spectral reflectance and algorithms
Meuleman, J., Bontsema, J., et al. (2002). An autonomous robot for for non-destructive chlorophyll assessment in higher plant leaves.
harvesting cucumbers in greenhouses. [Article]. Autonomous Robots, Journal of Plant Physiology, 160(3), pp. 271-282.
13(3), pp. 241-258. [23] Yam, K. L., & Papadakis, S. E. (2004). A simple digital imaging method
[4] Garcia, G. J., Pomares, J., & Torres, F. (2009). Automatic robotic tasks for measuring and analyzing color of food surfaces. Journal of Food
in unstructured environments using an image path tracker. Control Engineering, 61, pp. 137-142.
Engineering Practice, 17(5), pp. 597-608. [24] OceanControls. Retrieved 12/04/2010, from
[5] Kelly, R., Carelli, R., Nasisi, O., Kuchen, B., & Reyes, F. (2000). Stable www.oceancontrols.com.au
Visual Servoing of Camera-in-Hand Robotic Systems. Itransactions On [25] Sony Professional. Retrieved 05/06/2009, from www.pro.sony.eu
Mechatronics, 5(1), pp. 39 - 48. [26] Parallax Home. Retrieved 05/07/2010, from www.Parallax.com
[6] Reyes, JF; Chiang, LE., “Location And Classification Of Moving Fruits [27] Juckett, R. RGB color space conversion - Linear transformation of color.
In Real Time With A Single Colour Camera”, Chilean Journal Of Retrieved 14/8/2010, from http://www.ryanjuckett.com
Agricultural Research, Vol. 69, 2009, pp. 179-187
[28] Lindbloom, B. J. (2010). Retrieved 15/8/2010, from
[7] Scarfe, A. J., Flemmer, R. C., Bakker, H. H., & Flemmer, C. L. (2009). http://www.brucelindbloom.com/
Development of An Autonomous Kiwifruit Picking Robot. Proceedings
[29] Fleming, P. J., & Wallace, J. J. (1986). How not to lie with statistics:
of the Fourth International Conference on Autonomous Robots and
The correct way to summarize the benchmark results. Communications
Agents, pp. 639-643.
of the ACM, 29(3).

View publication stats