2009 Second International Symposium on Knowledge Acquisition and Modeling

Review of Recent Machine-vision Technologies in Agriculture

1

Bin Ji 1,2, Weixing Zhu1, Bo Liu1, Changhua Ma1, Xianfeng Li1

School of Electrical and Information Engineering, Jiangsu University, Zhenjiang, Jiangsu, China
2 School of Computer Science, Anhui University of Technology, Maanshan, Anhui,China
e-mail: [email protected], [email protected]

Abstract This paper reviews the machine-vision

techniques available for image acquisition and their
processing-analysis in agricultural automation up to now
according to the essential base and the core work, focusing
on 2 types of image, i.e., visible image and infrared image.
Results from previous studies have shown two important
topics on agriculture application, i.e., visual navigation and
behaviors surveillance. Each topic introduces important
and major applications in respective subfields. Furthermore, each one expresses technical discussions,
accordingly. Proper combination of different kinds of
machine-vision element for different objects is a normal
research way. The accuracy, the robustness and the realtime reveal 3 major obstacles in applications. Further to
this, many blanks are not yet developed, and the level of
some results is preliminary. Thus, additional research is
needed to fully realize this potential.
Keywords- agricultural automation;
visual navigation; behavioral surveillance

machine-vision;

1. INTRODUCTION
The marked characteristic of modern agriculture is
driving agricultural production towards automatization
and intelligentization by applying scientific and technological achievements, widely. In past decade years, with
the rapid development of information technology, the
machine-vision has been progressing in the theory and
technology. Profiting from the decline of technical costs
and the abundance of technical means, machine-vision
technology has penetrated into the agriculture field. In
turn, the biological diversity in agriculture boosts the
diversification and development of the technology. In
this paper, according to the extent of agricultural
application, the technological classification is shown in
three topics, i.e.: nondestructive measurement, visual
navigation and behavioral surveillance.
Here, agriculture is the generalized agriculture,
involving farming, forestry, animal husbandry, sideline
production and fisheries. Then this document is
organized as follows: Section 2, a core of the paper,
outlines the basic technical components by focusing on
image acquisition and image processing-analysis; Section
3, another core of the paper, shows two topics covered by
the application and the analysis; finally, section 4
presents the conclusions.
2. MACHINE-VISION IN AGRICULTURE
Machine-vision in agriculture expresses briefly that
the machine simulates the human eye to complete the
work processes of the measurement and the evaluation in
agricultural production for replacing humans operations.
So far, no existing a fully sense of replacement, it is only
a partial sense or to some extent replacement, but its
estimation level had been higher than the human eye in
978-0-7695-3888-4/09 $26.00
$25.00 2009 IEEE
DOI 10.1109/KAM.2009.231

334
330

some occasions, for example, identification of pestattacked maize plants by using infrared images[1].
Based on the processing, machine-vision technology
is divided into three phases: image acquisition, image
processing-analysis, I/O control. By utilizing the three
phases of technological elements, research groups set
respectively their own objectives. The new research
achievements are being a year-on-year growth, but
mainly converging on the image processing analysis one.
I/O controls are of industrial components control, mostly
being mature, not redesigning. Therefore, component
suppliers are obliged to provide some indispensable
technical support. Although the knowledge in machinevision systems is essential, the time and the effort for
them are negligible in the whole process of the
development. Therefore, this section primarily introduces
two parts, i.e. image acquisition and image analysis.
2.1 Image Acquisition
For the study objective, designing appropriate image
acquisition system, the basis of machine-vision system,
is essential. The purpose is acquisition of the
corresponding-quality images to satisfy the requirements
of the analysis.
Hardware technologies represented by CPU, DSP and
large-scale programmable logic device have been progressing with leaps and bounds. And in computing speed,
price, power and so on, the advantages are more and
more obvious. So, these make the image processing be
more and more widely applied in agriculture. The
technologies of machine-vision become rich. Since each
acquisition always has its own advantages and
limitations, confirmation of an acquisition largely
depends on the researchers own understanding the
studied object and expecting the image quality.
The light source, an important component of a
machine-vision system, must be emphasized in the
design of the system. To some extent, the choice and
design of light sources is generally the key to any
successful machine-vision system.
At present, the acquisition techniques of machinevision in agriculture mostly show in visible light
according to spectral bands. And applying some
invisible lights such as infrared are also gradually rising.
Subject to expensive imaging and technical instability,
the proportion of applying others invisible light is small.
But its unparalleled advantages are being valued by
more and more researchers.
Visible image: Now the types of image acquisition
equipment are many enough for the researchers choice.
Resources of visible images are in rich, but quite
complex. The imaging conditions are comprised of both
cases, i.e. indoor and outdoor, of which, the most
complex one is outdoor conditions that is divided into

target is performed after the highlighting. The main

methods of target extraction (binary): deletion of
pseudo-target,
adaptive
threshold,
progressive
clustering, multi-information fusion, etc.
The image processing in machine-vision is mostly
the detection and identification to the specific target.
The methods are different with different targets. The key
is: confirming the features extractable, preprocessing for
highlighting the target information, and taking the target
out by flexible methods. The nontarget is restrained till
the target has been highlighted [6].
Image analysis is an extension of target detection.
This is a process of express knowledge for the region
interested. The extracting valuable features are ready for
pattern recognition. Pattern recognition is capable of
abstracting characteristics and inferring the categories.
Its system contains several stages: representation,
adaptation, generalization or inference and evaluation
[7].

sunny, overcast, cloudy, evening and so on. Indoor

conditions are divided into natural light, incandescent
one, fluorescent one and others. Under different
conditions images show different characteristics. Such
as the obvious shadows of objects observed under the
sunny or the incandescent light, in addition the image
brightness is fluctuation with illumination varying.
These two shortcomings inspire many researchers
appetite for studying various solutions under different
conditions [2] for improving final result of the pattern
recognition.
Infrared image: Infrared equipment is separated into
the active and the passive according to whether loading
infrared light-emitting source or not. The passive is
applicable to imaging at close range. Conversely, the
active is applicable to long distance. The advantages of
infrared imaging display good antijamming, all-weather
work and so on. The image contains the target shape,
size and posture, and shows grayscale distribution. Its
signal-to-noise ratio is smaller. Meanwhile, the image
will have a greater change in large temperature
fluctuations.Active thermal infrared image may measure
the leaf water content. Passive one may estimate the
variable quantity associated with the stress of crop water
transpiration [3], and detect field crop disease [4]. Using
near-infrared spectroscopy and computer imaging
technology can do measurement of crop growthinformation and nondestructive quality inspection of
agricultural products etc.At present, the development of
infrared images for processing intelligentization is far
from penetrating and comprehensive.

3. TWO HOT TOPICS ON MACHINE-VISION TECHNOLOGY IN AGRICULTURE

By the impact of machine-vision on agriculture over

the past decade, results from previous studies have
presented 3 hot topics, i.e., nondestructive measurement,
visual navigation and behaviors surveillance. Nondestructive measurement, related to the static images
operating, is oldest and widest among them [8]. But the
latter two ones, related to the intelligent video, are
getting hotter and hotter. Thus, they are focused here.
3.1 Visual navigations
Visual navigation of agricultural machinery is a
research hotspot of intelligent agricultural machinery
recently [9], [10]. In automatic planting, such as spraying
fertilizer and pesticide, harvesting crops, cultivating and
weeding, etc, visual navigation plays a major role in
intelligent agricultural vehicles and refining agricultural
production. Thus, it is significant and prospective to
agricultural operating.
An agricultural machine may sense the surrounding
itself by visual information sensors. The information
what a navigation system needs is derived from the
sensor. Therefore, valid information and real-time
processing are crucial to the navigation system.
Key Technologies and Progresses:
Crops are usually cultivated along parallel lines, thus,
studying machine-vision is in charge of identifying the
crop rows, i.e. detecting the crops line from the image
for creating the parameters of vehicle navigation[11],
[12]. The row identification and the vehicle posture
assessment often use Hough transform, prototype
matching, statistical properties analysis, and virtual
intersection point analysis etc. Then distinguishing the
postures of vehicle depends on the relations of
calibration camera data and perspective geometry.
Visual navigation methods about agriculture robots
are divided into two categories: artificial signs and
characteristics of working environment. Artificial signs
have advantages of marking the path identification

2.2 Image Processing-analysis

The image processing-analysis is the core of the
machine-vision.
General steps of image processing are: color space
transforming, filtering, graying, and edge detection, etc.
These meet the needs of image segmentation later.
Video is an image sequence timeline-based. And the
beginning of video analysis is still processing image
frames.
Solutions of the detection and recognition of target is
a central issue of image processing, i.e. obtaining the
features of the target and nontarget describable correctly.
First, it must be coincided to how to separate target from
nontarget automatically. Generally, there are two means:
one is extraction of the target itself directly, which is
suitable for the distinct differences between target and
nontarget, such as gazing at green of the plant as the
characteristics under the varying sun-light to extract the
plant from the mixture that contain the fallen leaves and
soil successfully[5]; The other one is extraction of
closed border from the surrounded target, which is
mainly used in the situation that the differences between
target and nontarget are small so that it needs to
highlight the features of the target. There are too many
methods on highlighting, such as highlighting color,
cumulative characteristics etc. In fact, the extracting

331
335

and behavior analysis. Designing a robust target

detection algorithm being applicable to various
surveillance conditions, i.e. not being affected by
outdoor factors, such as illumination and
weather, is one of goals what the researchers
seek. But these algorithms are difficult to avoid
complexity. And most of the algorithms like
background
subtraction,
adjacent
frame
difference and optical flow method etc. aim to a
kind of specific issues [18].
Target classification: The key of target
classification is extracting features and designing
classifiers. The features express space-time
characteristics. Spatial characteristics are things
like shape, size, location [19] and so on related to
the target in a single-frame image. Time
characteristics are associated with tracking, such
as fluctuation in scale and speed and so on.
About designing classifier, common classifiers
are Bayesian, Neural network, as well as newly
SVM, etc. The SVM algorithms has been
widespread concerned, as it is competent for
small sample classification.
Target Tracking: The process of tracking of
moving target is: the characteristics uniquely
expressing the target need be found according to
difference between the targets and the
surrounding, and then the target locations need
be found by best matching the characteristics in
the follow-up frames. It is not only the main date
source of target detection and identification, but
also ready for target motion analysis and scene
analysis. Among many tracking algorithms, the
fundamental differences present in features
designing-choice and algorithms searching
match them. Several common tracking
algorithms are: feature-based approach [20], 2-3
dimension-based model method [21, [22], active
contour model-based method [37] etc.
Analysis of video content: After detecting and
tracking, video content needs be autoanalyzed
and understood, and given some appropriate
judgments. It is in the advance level of the
monitoring technology, and its the key of
realizing surveillance. In practical applications,
video content analysis includes target
identification, abnormality detection, and video
content description and so on.
Key Technologies and Progresses :
Now, video analysis of animal behavior is at the
primary stage: the research subjects category is seldom,
most objects are experimental rats for researching in
stress reaction in medicine. Since 1990 a system on
behavioral science analysis [23] has been developed to
apply in Neuroscience, pharmacology and toxicology by
Noldus Company (Netherlands). It is a remarkable
achievement. The chairman is also an expert [24] in
insect behavior by using machine-vision. At present,
animal categories in the research are limited to cattle,
pigs, fish and poultry, and the achievements are not yet
rich. The most studying purpose is for animal welfare
[25]. Despite the application in agriculture is not yet

simply, reliable, low-cost and achieving easily image

processing. As demanding highly operating environment
of the robot, it generally occurs in indoors with well
ground conditions. The other method is based on the
spatial features distribution round the robot. The later
have been applied to spray herbicides, fertilize and clean
pig house [13], [14].
In recent years, multi-information fusion technology has
been introduced to robot navigation. Multi-sensor may
attain redundant, complementary and real-time spatial
information. For solving navigating ineffectively problems, Multi-information fusion navigation converged
with GPS can improve effectively accuracy and
reliability, even realize full sense self-driving
completely. The visual navigation technology is
introduced in more detail [13].
Discussion :
Most visual navigation operation is under outdoor
natural light. And the objects in image are scattered,
irregular and complicated generally. The efficiency and
the robustness of algorithm are considered to satisfy
real-time image processing in complex and volatile
condition. Hough transform as a conventional algorithm
inspecting parameters of straight line is suited to image
inspection in unstructured area like crop-land for
tolerating noise and incompleteness. As a global
detection algorithm with expensive computation, it is
unfit for real-time agricultural operation. By combining
with the dynamic window technology, the algorithm
may greatly speed up the computing. In course of
searching for approaches on image description and
processing, if the methods like wavelet transform, multiscale theory and modern time series can be introduced
into the feature recognition linked with random
phenomena in agricultural engineering, new breakthroughs may be occur for solving detection various
edges of farmland images effectively. These would be
the studying direction in the next few years.
3.2 Behavioral surveillance
In agriculture, behavioral surveillance is: extracting
relevant visual information from the sequence images of
video, expressing them in appropriate methods, and then
interpreting them for identifying and studying animal
behavior. With intelligent video surveillance technology
rising in recent years [15], it is first applied to the field of
animal behavior and Neuroscience [16], [17]. It is
significant to intelligent monitoring in extensive breeding
industry.
As to intelligent video, after obtaining video
sequences by camera, firstly image quality needs be
improved through the gray or the super-resolution
restoration techniques. Then the target in the scene needs
be detected, classified and traced. Finally, video content
needs be understudied, such as inspecting anomaly in
scenes and identifying target. Specific process is as
follows:
Target detection: Its the start of the visual
monitoring technology, and a base of the
following process, such as target classification

332
336

The research of the machine-vision can not be

separated from the objective. Researching machinevision combined with the agricultural fields will also
promote the development of machine-vision itself. The
philosophy of researching machine-vision in agriculture
is: facing the diversity of agricultural objects, there is no
unique feature model and analysis way; on the other
hand, because there is no uniqueness on them, in turn, it
enriches the research methods effectively. All in all, the
criterion about evaluating the approach depends on the
special object itself and the effectiveness of the results.
Machine-vision technology in agriculture application
may be improved with the development of its own.
Meanwhile, many special issues do not still get the
settlement in the machine-vision technology because of
the richness and the complexity to agricultural objects.
Therefore, there are space and opportunities in future
development.
At present, most researches are to static image
processing. But in agriculture, there are many cases
which need real-time video processing, such as the
navigation of the agricultural vehicle and the surveillance
of the animal behavior. Therefore, it needs higher
demands to the real-time recognition algorithm. Using
support vector machines and manifold theory in pattern
recognition may have a light future. In brief, a truly
valuable agricultural machine-vision system must
possess real-time, robustness and accuracy. In the actual
research, we should consider various factors to make a
systematic research in this field combined with advanced
knowledge of the artificial intelligence and automatic
control disciplines.

prevalent, it is concerned more and more. Studies of

several categories are as follows:
On cattle: vision-based continuous quantication of lameness may analyze the track way in
cows locomotion [26]; a study compares video
with direct observation methods for measuring
oral behavior [27]; an automatic real-time
monitoring technique allows identifying the
locomotion and posturing behavior of pregnant
cows prior to calving [28].etc.
On the pig: a visual system is developed for
analyzing the growth of pigs [18]; by using
visual image analysis, pig growth is described
[29], [30]; A real-time computer vision is used to
assess and control thermal comfort for grouphoused pig [31], [32], etc.
On the fish: a video-based movement analysis
system is applied to quantify behavioral stress
responses of fish [33]; a video analysis is help
assessing vertical fish distribution in aquaculture
tanks [34].
Discussion:
The intelligent video mainly remains in the study of
a single object and the preset environment [35], mostly
using a single camera. Designing the animals moving
parameters is the basis of animal behavior research. The
research needs predefining behavior types under the
guide of animal ethology.
Incident detection methods mainly show the modelbased and the classifier-based. Event modeling contains
template-based method and statistics-based one. The
former is: establishing one or more templates for each
target sample, and then performing a matching process
as recognition between characteristic data and the
templates. For video analysis on animals behavior,
someone suggest a basic behavior unit segmentation
from visual data [36]. As solutions of inconsistent
movement speed in sequences, dynamic time warping
method is usually suggested. Statistical-based method is
widely used in pattern recognition of dynamic system.
Of which, Hidden Markov Model method is the most
representative one.
Currently, behavioral surveillance of intelligent
video applying in agriculture is still seldom, namely, at
the exploration stage. Thus it offers great potential for
development. The study is a challenging and significant
task in theory and practice. There are still three major
difficulties: robustness, accuracy, real-time. Therefore, a
large amount of works need be done to realize the
intelligent monitoring.

ACKNOWLEDGMENT
The authors wish to thank Professor Hanping Mao
and two anonymous referees provided helpful comments
on the manuscript from Jiangsu University (China). This
study was supported by NNSF of China (No. 30871449).
REFERENCES
[1]

[2]

[3]

[4]

4. CONCLUSIONS
This paper mainly selects the related research results
over the past decade. Because the periodical possesses a
better theoretical integrity, they are paid more attention in
collecting literature. On the view of the basic theory of
the machine-vision and the agricultural application, here
addresses the technical base and the application hot-spot
from these studies.

[5]

[6]

333
337

R.S.Zandonadi, F.A.C.Pinto, D.G.Sena Jnior, D.M.Queiroz,

P.A.Viana, E.C.Mantovani, " Identication of lesser cornstalk
borer-attacked maize plants using infrared Images".Biosystems
Engineering, vol. 91, Aug. 2005, pp.433-439, doi:10.1016/j.
biosystemseng.2005.05.002
H.T.Sgaard, H.J.Olsen, "Determination of crop rows by image
analysis without segmentation," Computers and Electronics in
Agriculture, vol.38, Feb. 2003, pp.141-158, doi:10.1016/S01681699(02)00140-0
M. Stoll, H.R. Schultz, B.Berkelmann-Loehnertz, "Thermal
sensitivity of grapevine leaves affected by Plasmopara viticola
and water stress," Vitis, vol. 47, 2008, pp.133-134
M.Stoll, H.R.Schultz, G.Baecker, B.Berkelmann-Loehnertz,
"Early pathogen detection under different water status and the
assessment of spray application in vineyards through the use of
thermal imagery,"Precision agriculture,vol.9,2008,pp.407-417,
doi:10.1007 / s11119-008-9084-y
D.S. Shrestha, B.L.Steward, "Automatic corn plant population
measurement using machine-vision,"Transactions of the ASABE,
vol.46, Mar-Apr. 2003, pp.559-565
J.Subbiah,N.Ray,G.A.Kranzler, S.T.Acton, "Computer vision
segmentation of the longissimus dorsi for beef quality grading,"
Transactions of the ASABE,vol.47, July-Aug. 2004, pp.12611268

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]

[21]

[22]

R. P.W.Duin, E. Pekalska, "The science of pattern recognition.

achievements and perspectives", in &hallenges for &omputational
Intelligence, Studies in Computational Intelligence Series, W.
Duch and J. Ma ndziuk, Eds.Springer, 2007, pp. 221-259,doi:
10.1007/978-3-540-71984-7_10
G.P.Moreda, J.Ortiz-Canavate, F.J.Garcia-Ramos, M.RuizAltisent, "Nondestructive technologies for fruit and vegetable
size determination-a review,"Journal of Food Engineering,vol.92,
May.2009,pp.119-136,doi:10.1016/j.jfoodeng.2008.11.004
D.C.Slaughter, D.K. Giles, D.Downey, "Autonomous robotic
weed control systems:A review,"Computers and Electronics in
Agriculture, vol.61,Apr. 2008,pp.63-78, doi:10.1016/j.compag.
2007.05.008
T.Torii, T.Teshima, T. Okamoto, k.Imou,"Vision based navigation of a boom sprayer,"Journal of the JSAM, vol.65,May. 2003,
pp.70-75
B.Astrand, A.J.Baerveldt,"A vision based row-following system
for agricultural field machinery," MECHATRONICS,vol.15,
2005,pp.251-269, doi:10.1016/j.mechatronics.2004.05.005
N.D.Tillett, T. Hague,S.J.Miles, "A field assessment of a
potential method for weed and crop apping on the basis of crop
planting geometry,"Computers and Electronics in Agriculture,
vol.32, Oct. 2001,pp.229-246, doi:10.1016/S0168-1699(01)
00167-3
S.T.Lu, X.Z.Wang, "Review of automatic guidance technologies
applied in agricultural vehicles," Proceedings of 2008 IAEA,
Zhenjiang,China,2008,pp.245-249
N.A.Andersen,I.D.Braithwaite,M.Blanke,T.Sorensen, Combining
a novel computer vision sensor with a cleaning robot to achieve
autonomous pig house cleaning,"Proceedings of the 44th IEEE
CDC-ECC '05, Seville, Spain,vol.2005, 2005, pp.8331-8336
N.Haering, P.L.Venetianer, A.Lipton,"The evolution of video
surveilance:an overview," Machine vision and applications,
vol.19, Oct.2008,pp.279-299, doi: 10.1007/s00138-008-0152-0
N.M.Lind,M.Vinther,R.P.Hemmingsen,A.K.Hansen, "Validation
of a digital video tracking system for recording pig locomotor
behaviour," Journal of neuroscience methods,vol.143, Apr.
2005,pp.123-132, doi:10.1016/j.jneumeth.2004.09.019
A.E.Poole,R.A.Stillman,K.J.Norris,"A video-based method for
measuring small-scale animal movement, "Animal Behavior,
vol.72, part5, Nov. 2006, pp.1205-1212, doi:10.1016/j.anbehav.
2006.02.019
W.M.Hu, T.N.Tan, M.B.Steve, "A survey on visual surveillance
of object motion and behaviors," IEEE Transactions on Systems,
Man and Cybernetics Part C: Applications and Reviews, vol.34,
Aug 2004, pp.334-352,doi: 10.1109/TSMCC.2004.829274
A.B.L.Tort,W.P.Neto,O.B.Amaral, V.Kazlauckas,D.O.Souza, "A
simple webcam-based approach for the measurement of rodent
locomotion and other behavioural parameters,"Journal of
neuroscience methods,vol. 157,Oct. 2006,pp.91-97,doi:10.1016/j.
jneumeth.2006.04.005
.Cangar, T. Leroy, M.Guarino, E. Vranken,R. Fallon,J.
Lenehan, et al, "Auomatic real-time monitoring of locomotion
and posture behaviour of pregnant cows prior to calving using
online image analysis," Compurters and electronics in
agriculture, vol.64, Nov.2008,pp.53-60,doi:10.1016/j.compag.
2008. 05.014
E.Marchand,P.Bouthemy,F.Chaumette,"A 2D-3D model-based
approach to real-time visual tracking,"Image and vision
computing, vol.19,Nov. 2001, pp.941-955, doi:10.1016/S02628856(01)00054-3
J.H.Wu, R.Tillett,
N.McFarlane, X.Y.Ju, J. P.Siebert,
P.Schofield, "Extracting the three-dimensional shape of live pigs

[23]

[24]

[25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

[33]

[34]

[35]

[36]

[37]

334
338

using stereo photo-grammetry," Computers and Electronics in

Agriculture. vol.44, Sep.2004,pp.203-222, doi:10.1016/j.compag.
2004.05.003
A.J.Spink,R.A.J.Tegelenbosch,M.O.S.Buma,L.P.J.J.Noldus,"The
Etho-Vision video tracking system - A tool for behavioral
phenotyping of transgenic mice,"Physiology & behavior, vol.73,
Sp.lss.si, Aug. 2001, pp. 731-744,doi: 10.1016/S0031-9384(01)
00530-3
L.P.J.J.Noldus, A.J.Spink, R.A.J. Tegelenbosch,"Computerised
video tracking, movement analysis and behaviour recognition in
insects," Compurters and electronics in agriculture, vol.35,Aug.
2002,pp.201-227, doi:10.1016/S0168-1699(02)00019-4
C.T.Whittemore, C.P.Schofield,"A case for size and shape
scaling for understanding nutrient use in breeding sows and
growing pigs, "Livestock production science,vol.65, Aug.
2000,pp.203-208, doi: 10.1016/S0301-6226(99)00136-0
X.Y.Songa, T.Leroya,E.Vrankena, W.Maertens, B.Sonck,
D.Berckmans, "Automatic detection of lameness in dairy cattle
Vision-based trackway analysis in cows locomotion,"
Computers and Electronics in Agriculture, vol.64, Nov. 2008,
pp.39-44, doi: 10.1016/j.compag.2008. 05.016
M. V. Tosi, V. Ferrante, S. Mattiello, E. Canali, M. Verga,
Comparison of video and direct observation methods for
measuring oral behaviourin veal calves, ITAL.J.ANIM. SCI.
,vol.5,2006,pp.19-27
R.P.White, C.P.Schofield, D.M.Green, D.J.Parsons, C.T.
Whittemore, "The effectiveness of a visual image analysis (VIA)
system for monitoring the performance of growing/finishing
pigs,"Animal science, vol.78,Part 3,Jun. 2004, pp.73-83
A.B.Doeschl-Wilson, C.T.Whittemore,P.W.Knap, C.P.Schofield,
"Using visual image analysis to describe pig growth in terms of
size and shape,"Animal science, vol.79,part 3,Dec. 2004,pp.415427
K.Kollis,C.S.Phang,T.M.Banhazi, S.J.Searle, "Weight estimation
using image analysis and statistical modelling: A preliminary
study,"Applied engineering agriculture,vol.23,Jan.2007,pp.91-96
D.J.Parsons, D.M.Green , C.P.Schofield ,C.T.Whittemore, "Realtime control of pig growth through an integrated management
system," Biosystems engineering, vol.96,Feb 2007,pp.257-266,
doi:10.1016/j.biosystemseng.2006.10.013
B.Shao, H.W. Xin,"A real-time computer vision assessment and
control of thermal comfort for group-housed pigs,"Computers
and electronics in agriculture,vol.62,Jun. 2008,pp.15-21,
doi:10.1016/ j.compag.2007.09. 006
A.S.Kane,J.D.Salierno,G.T.Gipson ,T.C.A. Molteno, C.Hunter,"
A video-based movement analysis system to quantify behavioral
stress responses of fish," Water Research, vol.38, 2004,
pp.3993-4001, doi: 10.1016/j.watres.2004.06.028
L.H.Stien, S.Brafland,I.Austevollb,F.Oppedal, T.S.Kristiansen,
"A video analysis procedure for assessing vertical fish
distribution in aquaculture tanks,"Aquacultural engineering,
vol.37,Sep.2007,pp.115-124,doi: 10.1016/j.aquaeng.2007.03.002
Y.J.Chen , Y.C.Li, K.N.Huang ,S.L. Jen, M.S.Young, "Video
tracking algorithm of long-term experiment using stand-alone
recording system,"Review of Scientific Instruments, vol.79, Aug.
2008, 085108, doi:10.1063/1.2976035
X.W. Xue, T.C.Henderson,"Exploration of The Vector Fusion
Method for Basic Behavior Unit Segmentation from Visual
Data,"IEEE International Conference on MFI, Heidelberg,
Germany, 2006,pp.122-126, doi: 10.1109/MFI.2006.265674
W.Kim, J.J.Lee," Object tracking based on the modular active
shape model," MECHATRONICS,vol.15,Apr. 2005,pp.371-402,
doi:10.1016/ j.mechatronics.2004.09.002