ArticlePDF Available

Abstract and Figures

Specialty crops, like flowers, herbs, and vegetables, generally do not have an adequate spectrum of herbicide chemistries to control weeds and have been dependent on hand weeding to achieve commercially acceptable weed control. However, labor shortages have led to higher costs for hand weeding. There is a need to develop labor-saving technologies for weed control in specialty crops if production costs are to be contained. Machine vision technology, together with data processors, have been developed to enable commercial machines to recognize crop row patterns and control automated devices that perform tasks such as removal of intrarow weeds, as well as to thin crops to desired stands. The commercial machine vision systems depend upon a size difference between the crops and weeds and/or the regular crop row pattern to enable the system to recognize crop plants and control surrounding weeds. However, where weeds are large or the weed population is very dense, then current machine vision systems cannot effectively differentiate weeds from crops. Commercially available automated weeders and thinners today depend upon cultivators or directed sprayers to control weeds. Weed control actuators on future models may use abrasion with sand blown in an air stream or heating with flaming devices to kill weeds. Future weed control strategies will likely require adaptation of the crops to automated weed removal equipment. One example would be changes in crop row patterns and spacing to facilitate cultivation in two directions. Chemical company consolidation continues to reduce the number of companies searching for new herbicides; increasing costs to develop new herbicides and price competition from existing products suggest that the downward trend in new herbicide development will continue. In contrast, automated weed removal equipment continues to improve and become more effective. Available at:
Content may be subject to copyright.
BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, research
libraries, and research funders in the common goal of maximizing access to critical research.
Technology for Automation of Weed Control in Specialty Crops
Author(s): Steven A. Fennimore, David C. Slaughter, Mark C. Siemens, Ramon G. Leon, and Mazin N.
Source: Weed Technology, 30(4):823-837.
Published By: Weed Science Society of America
BioOne ( is a nonprofit, online aggregation of core research in the biological, ecological, and
environmental sciences. BioOne provides a sustainable online platform for over 170 journals and books published
by nonprofit societies, associations, museums, institutions, and presses.
Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance of
BioOne’s Terms of Use, available at
Usage of BioOne content is strictly limited to personal, educational, and non-commercial use. Commercial inquiries
or rights and permissions requests should be directed to the individual publisher as copyright holder.
Weed Technology 2016 30:823–837
Technology for Automation of Weed Control in Specialty Crops
Steven A. Fennimore, David C. Slaughter, Mark C. Siemens, Ramon G. Leon, and Mazin N. Saber*
Specialty crops, like flowers, herbs, and vegetables, generally do not have an adequate spectrum of
herbicide chemistries to control weeds and have been dependent on hand weeding to achieve
commercially acceptable weed control. However, labor shortages have led to higher costs for hand
weeding. There is a need to develop labor-saving technologies for weed control in specialty crops if
production costs are to be contained. Machine vision technology, together with data processors, have
been developed to enable commercial machines to recognize crop row patterns and control
automated devices that perform tasks such as removal of intrarow weeds, as well as to thin crops to
desired stands. The commercial machine vision systems depend upon a size difference between the
crops and weeds and/or the regular crop row pattern to enable the system to recognize crop plants
and control surrounding weeds. However, where weeds are large or the weed population is very
dense, then current machine vision systems cannot effectively differentiate weeds from crops.
Commercially available automated weeders and thinners today depend upon cultivators or directed
sprayers to control weeds. Weed control actuators on future models may use abrasion with sand
blown in an air stream or heating with flaming devices to kill weeds. Future weed control strategies
will likely require adaptation of the crops to automated weed removal equipment. One example
would be changes in crop row patterns and spacing to facilitate cultivation in two directions.
Chemical company consolidation continues to reduce the number of companies searching for new
herbicides; increasing costs to develop new herbicides and price competition from existing products
suggest that the downward trend in new herbicide development will continue. In contrast, automated
weed removal equipment continues to improve and become more effective.
Key words: Automation, integrated weed management, intelligent cultivator, intrarow weed
control, mechanization, robotic weeding, vegetable crops
Los cultivos hort´
ıcolas de alto valor tales como flores, hierbas, y vegetales generalmente no tienen un espectro adecuado de
ımicos herbicidas para el control de malezas y han sido dependientes de la deshierba manual para alcanzar un control de
malezas comercialmente aceptable. Sin embargo, la escasez de mano de obra ha provocado el incremento en los costos de la
deshierba manual. Si se pretende contener los costos de producci´
on, existe una necesidad de desarrollar tecnolog´
alternativas a la mano de obra para el control de malezas en cultivos hort´
ıcolas de alto valor. La tecnolog´
ıa de ma
´quinas de
on, combinada con procesadores de datos, ha sido desarrollada para hacer posible que ma
´quinas comerciales puedan
reconocer los patrones de siembra en hileras del cultivo y a la vez controlar equipos automatizados que pueden desempe˜
labores tales como la remoci´
on de malezas en la hilera de siembra, o ralear la densidad de siembra del cultivo. Los sistemas
de ma
´quinas de visi´
on comerciales dependen de la diferencia entre el tama˜
no del cultivo y el de las malezas y/o de la
regularidad del patr´
on de distribuci´
on del cultivo para que el sistema pueda reconocer las plantas del cultivo y las malezas a
su alrededor. Sin embargo, donde las malezas son grandes o la poblaci´
on de malezas es muy densa, los sistemas de
´quinas de visi´
on actuales no pueden diferenciar efectivamente entre las malezas y los cultivos. Los equipos
automatizados de deshierba disponibles comercialmente hoy en d´
ıa dependen de cultivadores o aspersores dirigidos para
controlar malezas. Los equipos de acci´
on para el control de malezas en modelos futuros podr´
ıan usar abrasi´
on con aspersi´
de arena con aire a presi´
on o calor con equipos con llamas de fuego para matar las malezas. Las estrategias de control de
malezas en el futuro probablemente requerira
´n la adaptaci´
on de los cultivos al equipo automatizado de remoci´
on de
malezas. Un ejemplo de esto ser´
ıa el cambio de patrones de siembra y distancias entre hileras del cultivo para facilitar la
labranza en dos direcciones. La consolidaci´
on de compa˜
ıas qu´
ımicas contin´
ua reduciendo el n´
umero de compa˜
ıas que
´n buscando nuevos herbicidas. Adema
´s, el incremento en los costos de desarrollar nuevos herbicidas y el precio de la
competencia a partir de productos existentes sugiere que la tendencia decreciente en el desarrollo de nuevos herbicidas
´. En contraste, equipos automatizados de remoci´
on de malezas contin´
uan mejorando y haci´
endose ma
´s efectivos.
DOI: 10.1614/WT-D-16-00070.1
* Extension Specialist, Department of Plant Sciences, University of California, Davis, Salinas, CA 93905; Professor, Department of
Biological and Agricultural Engineering, University of California, Davis, Davis, CA 95616; Associate Professor, Department of
Agricultural and Biosystems Engineering, University of Arizona, Tucson, AZ 85721; Assistant Professor, West Florida Research and
Education Center and Agronomy Department, University of Florida, Jay, FL 32565; Postdoctoral Research Associate, University of
Arizona, Yuma Agricultural Center, AZ 85364. Corresponding author’s E-mail:
Fennimore et al.: Weed control automation in specialty crops 823
Automated weed control technology is available
commercially and is being used in vegetable crops
like broccoli (Brassica oleracea L.) and lettuce
(Lactuca sativa L.) (Lati et al. 2016). Because of
the limited availability of effective herbicides, and
high labor used in the production of specialty crops
(flower, fruit, herb, vegetable, and other horticul-
tural crops), these crops are most likely to benefit
from weed control automation due to reduced labor
inputs required for hand weeding (U.S. Department
of Agriculture [USDA] 2016). As a group, the
highly varied plant architectures and production
systems of flower, fruit, herb, vegetable, and other
horticultural crops make it challenging if not
impractical to develop a one-size-fits-all automatic
weeding machine for specialty crops. Weed removal
technology developed in specialty crops may prove
useful for application in major crops, albeit with
required modification for canopy coverage, planting
density, as well as operation speed and cost per
hectare. Within the last decade, new technologies
for crop thinning and weed removal have been
developed and commercialized for specialty crops,
and there will likely be many more future advances
in machine learning and development of smart
machines for agriculture.
Herbicides have greatly reduced agricultural
production costs as well as contributed toward
increasing crop yields (Gianessi and Reigner 2007).
A major stimulus for introduction of herbicides was
the loss of farm labor for hand weeding and
increasing farm labor costs during the 1940s and
1950s. This led to the weed management revolution
based on conventional herbicides during the 1960s
and 1970s, followed by the introduction of
glyphosate-resistant crops during the 1990s and
early 2000s (Duke 2012; Shaner 2000). Not all
crops participated equally in the dramatic increases
in efficiency that conventional herbicides and
glyphosate-resistant crops brought to the market.
Specialty crops, especially flower, berry, and
vegetable crops, do not have an adequate spectrum
of herbicide chemistries to provide commercially
acceptable weed control by themselves; therefore,
hand weeding is also needed to protect many
specialty crops (Fennimore and Doohan 2008).
Hand weeding, cultivation, and cultural methods of
weed management are especially important in
organic crops where organic-compliant herbicides
are expensive and seldom used in commercial
agriculture (Boyd et al. 2006). Because of stricter
immigration policies and competition for laborers
from nonagricultural sectors, the United States is
again finding that farm labor supply does not meet
demand, and there is a need for alternatives to hand
weeding (Taylor et al. 2012). Rapid industrializa-
tion in Mexico with accompanying employment
opportunities and farm labor demands in that
country have resulted in less farm labor available in
the United States. This paper will explore possible
alternatives using automation technology to im-
prove weed control systems in specialty crops.
Commercial development of new herbicide active
ingredients has greatly slowed in the past decade. As
recently as the 1970s and 1980s, several new active
ingredients were introduced annually (Duke 2012).
Currently, new herbicide active ingredients may
cost $240 to $300 million from discovery to launch
(Castello et al. 2015; Duke 2012; Kraehmer et al.
2014; Lamberth et al. 2013; R¨
uegg et al. 2007).
High development costs will likely limit the number
of new herbicide active ingredients in the pipeline,
particularly for specialty crops where the number of
hectares is low compared to agronomic crops.
During the period 1970 to 1991, 15 new herbicide
modes-of-action were introduced in Europe, but
none after 1991 (R¨
uegg et al. 2007). During the
period 1980 to 2009, 137 new herbicides were
introduced worldwide (Duke 2012; Kraehmer et al.
2014). However, in the period 2010 to 2014 only
four new herbicides were introduced (Jeschke
2016). Perhaps the increasing difficulties in man-
aging weeds resistant to glyphosate and other
herbicides will stimulate some of the agricultural
chemical companies to reinvigorate their develop-
ment efforts (Duke 2012; Shaner 2000). However,
the much higher development costs for new
chemistries and the fact that there are fewer major
agricultural chemical companies today looking for
new herbicide chemistry, suggests the downward
trend of new herbicide introductions will continue
(Duke 2012). Also contributing to the lack of new
herbicide chemistry is the availability of established
herbicides already on the market and the intense
competition among agricultural chemical compa-
nies for market share, which holds down prices and
must be considered before making a decision to
invest in new products (R¨
uegg et al. 2007).
Automatic weed removal technology provides an
alternative path to weed control with reduced
824 Weed Technology 30, October–December 2016
dependence on both the agricultural chemical
industry and hand weeding. Recently, a growing
number of new European and U.S. companies with
expertise in machine vision, automation, mecha-
tronics (control systems and actuators), and robotics
have formed to address labor shortages in the
specialty crop industry. The likelihood that this
trend will continue is high, as the development of
automatic weed removal equipment can be much
less expensive than herbicides. For example, the
development costs for Steketee’s intelligent cultiva-
tor (IC) in The Netherlands, was approximately
$11 to $17 million (Leonard Mol, Steketee,
personal communication) which compares favor-
ably to the .$250 million required to develop a
new herbicide (R¨
uegg et al. 2007). New technolo-
gies that combine sensors and mechatronics with
existing weed removal technology like cultivation
knives and sprayers to create new tools to remove
weeds will be discussed.
Automation of weed control has two key aspects,
crop/weed detection and weed control actuators.
Automatic detection involves machine recognition
of key features of the plant and differentiation
between the crop and weed. Actuation takes two
approaches to weed automation; one approach is
chemical and the other mechanical (De Baerde-
maeker 2014). Here we will explore examples of
both of these approaches with automated lettuce
thinners used as an example of automated chemical
application and robotic intrarow cultivation as an
example of mechanical weed control automation.
Weed and Crop Detection
GPS Guidance. Because of the small size of the
crop at the optimum time in the plant life cycle for
weed control (Lanini and LeStrange 1991) precise
lateral guidance of the sensors and actuators is a
critical prerequisite for robust performance in
automated weed control systems, particularly those
based upon spray targeting of individual weeds
(Lamm et al. 2002). The most commonly used
technology for precise lateral guidance in specialty
crops is currently the real-time kinematic (RTK)
global positioning systems (GPS) technology for
autoguidance of the tractor. Most modern RTK-
GPS guidance systems are compatible with global
navigation satellite systems from multiple countries
to increase the total number of satellites available
and improve reliability. The U.S. Geological Survey
(USGS) reports that the accuracy of RTK-GPS
location measurement is 3 cm when determined in
relationship to a specific geodetic datum (USGS
2016). The second type of technology in use for
lateral guidance of automated weed control systems
is based on machine-vision crop row following
(Fennimore et al. 2010). Under low to moderate
weed infestation levels, machine vision guidance
systems can outperform RTK-GPS guidance sys-
tems because they are more accurate (Slaughter et al.
1999) and because they are typically set up to
control the lateral positioning of the cultivation
system directly, rather than indirectly through
control of the tractor. At high weed infestation
levels machine-vision guidance systems can become
unreliable. The convenience of having one guidance
system for all farming operations (e.g., tillage,
planting, harvesting) and the independence of the
performance of RTK-GPS from the weed densities/
size means RTK-GPS guidance system will likely
become the predominant technology for precise
lateral guidance of automated weed control plat-
Once the lateral position of the robotic system has
been established, the next task is to identify
individual plants within the crop row. The two
weed control mechanisms (hoeing vs. spray) led to
different requirements of the sensing approach to
plant identification. For robotic hoeing, the sensing
need is for accurate detection and localization of the
crop plants’ centroids (center of plant symmetry),
and weed detection is not generally needed. For
robotic weed spraying, assuming a nonselective
phytotoxic solution precisely sprayed on the target
weed avoiding the crop and soil, the sensing need is
for an accurate map of the location of the weed
foliage and here, mapping the crop foliage can be
important for minimizing accidental targeting of
the crop. Of the two cases, detecting and mapping
the crop plants’ centroids is the least challenging
and has the greatest number of sensing modalities
Three general noncontact sensing approaches to
automatic detection and mapping of the crop
plants’ centroids have been demonstrated. The first
sensing modality requires a systems approach to
crop mapping and the use of RTK-GPS both at
planting and during weed control. This concept was
developed by Upadhyaya et al. (2003, 2005) and
Fennimore et al.: Weed control automation in specialty crops 825
has been experimentally verified by Ehsani et al.
(2004), Nørremark et al. (2007, 2008), Sun et al.
(2010), and P´
erez-Ruiz et al. (2012a,b). In this
method, an RTK-GPS crop plant map of the seed
location, or transplant location is automatically
made at planting. The RTK-GPS crop location map
is later used by the robotic weed control system to
control the position of the robotic hoes automat-
ically, pulling the hoes away from the close-to-crop
zone when approaching a crop location in the map,
and then back into the row center to control
intrarow weeds between crop plants. This sensing
approach has the advantages of not requiring any
crop- or weed-specific knowledge, is less computa-
tionally intensive than machine-vision techniques,
has no optical elements that require protection, and
does not require any illumination source or control,
but requires a GPS-mapping planter or transplanter
and the availability of a high-quality RTK-GPS
signal both at planting and during weeding events.
Recent advances in RTK-GPS technology have
greatly reduced the component cost of RTK-GPS
equipment, suggesting that this technique may
become cost-competitive with other technologies
in the near future.
Electromagnetic Absorbance. The second sensing
modality that has been demonstrated is the use of
X-ray or gamma ray sensing technology for
detection of crop plants. The concept was originally
developed for lettuce harvesting using X-ray by
Lenker and Adrian (1971) and gamma ray by
Garrett and Talley (1969) and has been experimen-
tally verified for tomato (Solanum lycopersicum L.)
plant detection in the field for weed control
applications by Haff et al. (2011). The method is
based upon the absorbance of electromagnetic
energy and requires that the main stem in a central
leader crop architecture or central portion in a
rosette architecture crop plant have greater absor-
bance than the weeds, typically by the crop being
physically larger. As such, the sensing method is
most appropriate for weed control in a transplanted
crop, where the crop plants are older and typically
larger than the emerging weeds. Like the GPS
method described above, it is also used to control
the position of robotic weeding hoes into and out of
the close-to-crop zone. It is the least computation-
ally intensive noncontact method of crop detection
that has been demonstrated in specialty crops.
Because it utilizes a penetrating source of electro-
magnetic radiation (unlike visible or near-infrared
light-based methods), it is robust to the visual
occlusion of the crop plant center by either crop or
weed foliage, an advantage over machine vision
techniques. The primary disadvantage is the need to
protect workers from accidental exposure to the X-
ray or gamma ray source.
Machine Vision Detection. The third, and most
predominant sensing modality used in automatic
weed control are machine vision techniques. The
most powerful, and to date the only method capable
of robust, automated in-field discrimination of
individual plant species, is based upon hyperspectral
imaging. The hyperspectral imaging concept has
been demonstrated in the field in lettuce (Slaughter
et al. 2008) and tomato (Zhang and Slaughter
2011; Zhang et al. 2012a,b), with between-species
pixel-level recognition rates above 75% and crop vs.
weed discrimination rates above 90%. The tech-
nique uses machine learning to establish a spectral
pattern recognition classifier and when used as a
proximal sensor (as opposed to remote sensing via
aerial vehicles) with both controlled illumination
and a thermally stabilized camera, can distinguish
between closely related species such as tomato and
black nightshade (Solanum nigrum L.). Other
advantages beyond species identification capability
are that it is less computationally intensive than
shape-based pattern recognition, it is robust to
visual occlusion of the leaf margin (another
advantage over leaf shape recognition), and the
species recognition ability can be used to customize
the spray application of multiple herbicidal mate-
rials based upon weed species. Its disadvantages are
that it requires a multiseason calibration process
(Zhang et al. 2012a,b) and must be trained to
distinguish between closely related species. The
current cost of hyperspectral imaging camera
technology is much higher than traditional digital
color imaging hardware; however, recent develop-
ments in multispectral cameras for the rapidly
growing aerial remote sensing market suggest that
lower-cost multiwaveband cameras may be available
in the near future.
The most widely studied and only commercially
utilized method of sensing in existing robotic weed
control machines is based upon traditional 2D
machine vision techniques. Although a number of
advanced machine vision recognition techniques for
plants have been documented in the literature (e.g.,
826 Weed Technology 30, October–December 2016
Hearn 2009; Manh et al. 2001; Persson and
Astrand 2008; Søgaard 2005), their high level of
computational intensity and the predominant use of
serial processors in existing robotic weed control
machines limits their adoption. More commonly,
2D image processing approaches, based upon a
combination of plant detection (from soil) by color
or infrared to red light reflectance ratios, and crop
recognition (from weeds) by apparent plant size in
the 2D image and the 2D spatial planting pattern
along the crop row are used (e.g., ˚
Astrand and
Baerveldt 2005; Onyango and Marchant 2003;
Southall et al. 2002; Tillett et al. 2001, 2008).
These techniques are fairly effective during the early
portion of the growing season, when weed control is
most critical, before canopy closure has occurred
and the 2D silhouettes of the plants have not yet
merged. For robust performance they require a
uniform and well-established crop stand and a
relatively low weed density, and perform best when
applied to a transplanted crop where the crop plants
are larger than and more easily distinguished from
weeds. The technique is better suited for the control
of a robotic hoe than a sprayer, where the need for
accurate recognition or mapping of weeds growing
in close proximity to the crop is not necessary.
Weed Control Actuators
These are devices like cultivators that kill weeds
by uprooting, thermal weeders like flame, lasers or
steam that kill the weeds by destroying plant
membranes, abrasives that physically degrade weed
foliage, and mowers that cut weeds. Weed detection
systems and processors signal the actuator to control
the weed but not the crop.
Physical Weed Removal. There are four primary
methods of physical weed control that may be
automated: (1) mechanical cultivation; (2) thermal
weed control; (3) abrasion, i.e., bombardment with
air-propelled abrasive grits; and (4) mowing. All of
these methods have the advantage of being pest
control devices ‘‘which work by physical means to
control a pest’’ as defined by the U.S. Environ-
mental Protection Agency (USEPA), and do not
require registration (USEPA 2015). Weed control
devices also have the advantage of being organic-
compliant and therefore useful in organic settings.
With the exception of air-propelled abrasive grits,
all of these physical methods for weed removal have
been in use for decades. Automation is a means to
take a proven weed control device like a cultivator
knife, and combine it with intelligent technology to
create something very different—an intelligent
Intrarow Cultivation. Traditionally intrarow weeds
have only been controlled by hand weeding and
selective herbicides (Haar and Fennimore 2003).
Cultivator tools such as finger weeders and torsion
weeders are also used, but generally only remove
small weeds (Cloutier et al. 2007, van der Schans et
al. 2006). Growers and researchers have been
searching for a way to remove weeds from the
intra-row space mechanically without damaging the
crop (Cloutier et al. 2007; Fennimore et al. 2014;
Melander et al. 2015; Tillett et al. 2008; van der
Schans et al. 2006).
Two designs for intelligent intrarow cultivators
are currently on the market: the rotating disc design
from Tillett and Hague Technology Ltd. in the
United Kingdom and is being marketed as the
Robocrop ‘‘In Row’’ cultivator (Fennimore et al.
2014; O’Dogherty et al. 2007; Tillett et al. 2008)
and reciprocating knives that reach in and out of the
crop row with the use of machine vision guidance
(Melander et al. 2015). Both systems essentially do
the same thing, but by different mechanisms. The
Robocrop cultivator has a rotating disc controlled
by a vision system to detect the crop plant and align
the disc cutaway section with the crop plant (Figure
1). The disc rotational phase is altered as needed by
changing the speed of the hydraulic drive to align
the cutaway section with the crop plant to allow for
variation in crop spacing (Tillett et al. 2008).
Research showed that the Robocrop rotating
cultivator provided effective weed control and did
not injure or reduce yields in transplanted vegetable
crops compared to the standard cultivator treatment
(Fennimore et al. 2014). However, in seeded lettuce
where plant spacing is more variable, the Robocrop
injured the crop and caused 26% yield reductions,
primarily due to stand reduction.
Examples of reciprocating knife type intelligent
cultivators currently available for commercial use in
North America include the Robovator and Steketee
IC. The Robovator intelligent cultivator was also
evaluated in broccoli and lettuce in comparison to a
conventional interrow cultivator in California
(Figure 2). The Robovator reduced the weed
densities by 27 to 41% more than the standard
Fennimore et al.: Weed control automation in specialty crops 827
cultivator and hand-weeding times by 29 to 45%
(Lati et al. 2016). They also found that broccoli and
lettuce yields were not reduced by the Robovator.
Another commercially available intelligent cultiva-
tor is the Steketee IC cultivator (Figure 3).
Evaluations of the IC cultivator in lettuce showed
that the machine found the crop row readily and
detected the crop with near 100% accuracy early in
the crop cycle, but fell to 54% accuracy at crop
maturity (Hemming et al. 2011).
Intelligent cultivators were found to be a viable
alternative to hand weeding for vegetables grown in
Denmark (Melander et al. 2015). Robovator
intelligent cultivator was evaluated in onion [Allium
cepa var. cepa (L.)] and cabbage [Brassica oleracea
(L.) var. capitata (L.)] for intrarow weed removal in
comparison to a finger weeder (Melander et al.
2015). They did not see a large advantage for the
intelligent cultivator compared to the ‘‘nonintelli-
gent’’ finger weeder in terms of weed control and
crop tolerance.
Another concept for controlling intrarow weeds
mechanically is ‘‘stamping.’’ Stamping is where
weeds are pushed into the ground with the use of a
high-speed ‘‘ramming rod.’’ Michaels et al. (2015)
developed one such system for use in organic carrot
[Daucus carota (L.) var. sativa (Hoffm.)] produc-
tion. The stamping tool comprised of a cordless nail
gun mechanism, which forced a 1.0-cm-diameter
cylinder rod 30 cm into the soil. The stamping tool
Figure 1. Robocrop intrarow cultivator used on two celery
[Apium graveolens L. var. dulce (Mill.) DC.] beds at Salinas, CA
(top). The bottom photo is a close up of the intrarow cultivation.
(Color for this figure is available in the online version of this
Figure 2. Robovator intrarow cultivator equipped with
reciprocating knives, weeding lettuce on a 2-m-wide bed near
Santa Maria, CA. (Color for this figure is available in the online
version of this article.)
Figure 3. Steketee IC, equipped with reciprocating knives
operating in a fennel (Foeniculum vulgare Mill.) planting near
Castroville, CA. (Color for this figure is available in the online
version of this article.)
828 Weed Technology 30, October–December 2016
was attached to a robotic arm that allowed the
device to move in the X–Y plane for precision
placement. The robot was equipped with two
cameras, one mounted in front of the vehicle for
detecting the general location of weed plants and
the second for precise positioning of the stamping
tool. Although actuator positioning was generally
better than 5 mm and stamping tool cycling speed
was 100 ms, the maximum work rate of the system
was only 1.75 weed/s
at a speed of 3.7 cm/s
Thermal Weed Control. Flame weed control, used in
organic and conventional cropping systems, is a
process of exposing plant tissues to flames coming
from a burner normally fueled by propane. Flaming
controls weeds by heating rather than burning plant
tissue. Propane burners can generate combustion
temperatures of up to 1,900 C, which raises the
temperature of the exposed plant tissues rapidly.
Heat injury results in destruction of plant mem-
branes, which results in loss of cell function, and
eventually the plants die or are severely weakened.
During the 1960s, flaming was widely used in the
United States for weed control in cotton (Gossypium
hirsutum L.), maize (Zea mays [Schrad.] Iltis),
sorghum [Sorghum bicolor (L.) Moench], soybean
[Glycine max (L.) Merr.], potato (Solanum tuber-
osum L.), and other crops. During the 1960s and
the 1970s, flaming was replaced by the use of
herbicides because of an increase in propane price
and the availability of less expensive herbicides.
Recently, concerns about herbicide impacts on the
environment have renewed interest in flame weed
control (Knezevic et al. in press).
The use of automated flamers for intrarow weed
removal remains an area to be explored (Nørremark
et al. 2009). A prototype automated flamer for
intrarow weed removal is under development in
Denmark by Poulsen Aps (Poulsen 2011). The basis
of this machine is serial application of a flame fueled
by propane. As the flamer passes over a weed, the
machine vision system turns on the burners so that a
specific weed will receive multiple exposures to the
heat. The machine vision system turns off the flame
as it passes over the crop. The potential for use of
precision steam application to control weeds either
by precision foliar application to control weeds or to
disinfect soil killing weed seed prior to crop
planting remains to be explored (Knezevic et al. in
Hot Oils. A robotic system for precision, pulsed-jet,
microdosing of high-temperature, organic, food-
grade oil for intrarow weed control in vegetable
crops was developed by Zhang et al. (2012b) at the
University of California, Davis. This robotic system
has two unique features. First, it was capable of
distinguishing tomato plants from weeds, including
black nightshade, with above 95% accuracy with the
use of an advanced sensing system that used
hyperspectral imaging and a Bayesian classifier
based upon machine learning for real-time species
recognition. Second, the robotic pulsed-jet, micro-
dosing system (the actuator) could target individual
weed leaves with a lethal dose of high-temperature
(160 C) oil with a 1-cm spray resolution. The target
application rate to weed leaf surface was 0.85 mg
of the thermal fluid, applied in 10-ms-pulsed
doses, which achieved above 93% weed control at
15 d after application with less than 3% of tomato
plants injured by the treatment. Commercial
development potential is unknown.
Laser Weeding. Weed control using laser technology
has been investigated by several researchers as a
nonchemical/organic weed method for manage-
ment. Mathiassen et al. (2006) studied the potential
of commercially available laser systems for control-
ling three weed species: common chickweed
(Stellaria media L.), scentless mayweed [Tripleur-
ospermum inodorum (L.) Sch. Bip.], and oilseed rape
(Brassica napus L.). Two different types of contin-
uous-wave diode lasers were used to target the apical
meristems of weeds at the cotyledon stage. Weed
control efficacy varied significantly and depended
on weed species, wavelength, exposure time, spot
size, and laser power. Of the systems tested, only the
5-W, 532-nm laser with 1.8-mm spot size config-
uration effectively controlled all three weed species.
Energy and exposure times required for this
configuration ranged from 1.3 to 9.9 J, with
corresponding exposure times of 250 to 2,000 ms.
Further research is needed to determine the efficacy
on a broader spectrum of weed species.
Kaierle et al. (2013) also experimented with using
lasers for nonchemical weed control. They investi-
gated four types of lasers at different wavelengths:
(10,600 nm), fiber (1,908 nm), diode (940
nm), and solid-state (532 nm). The lasers were used
to irradiate the apical meristem of redroot pigweed
(Amaranthus retroflexus L.) at the first true leaf stage
of growth. Three different laser positions and three
Fennimore et al.: Weed control automation in specialty crops 829
spot sizes 3.0, 4.2, and 6.0 mm were used to
investigate the impact of local or systemic irradia-
tion on plant damage. Minimum lethal doses of
energy ranged from 10 to 71 J weed
with the
lowest energy requirements resulting from use of the
laser. Laser positioning was found to be
important, as a damage model showed a required
energy increase of 1.3 J for every 1% loss in
positioning accuracy.
Abrasion. Air-propelled abrasive grit material has
been evaluated as a weed control device directed at
the base of corn, soybean, or vegetables (Forcella
2009, 2012, 2013; Wortman 2014, 2015). Similar
methods have been evaluated in the Netherlands
with pneumatic air blasting (van der Schans et al.
2006). Currently there is research to pair abrasion
with intelligent technology, but to the best of our
knowledge there are no published reports yet
(Manuel Perez Ruiz, personal communication).
Mowing. The main area where autonomous mowers
have been utilized is for mowing lawns (Melita et al.
2013). Traditionally, autonomous lawn mowers
have utilized a wire guidance system that sets the
boundaries for the area to be mowed. However, a
project to develop highly accurate GPS guidance for
autonomous lawn mowers is in progress in Europe
(Melita et al. 2013). Traditionally mowing has been
used for weed control in orchards, vineyards, and
pastures (Cloutier et al. 2007). There would appear
to be potential for use of autonomous mowers for
weed control in all of these areas. There has been
some research in Denmark to develop an autono-
mous mower for weed control in Christmas tree
plantations (Have et al. 2005). It would appear that
the potential exists for further development of
autonomous mowing vehicles for several applica-
tions in orchards, vineyards, and pastures.
Precision-Spray Application. Precision spraying is
where small volumes of herbicidal spray are directed
to areas near the crop plant to control weeds.
Precision-spray assemblies are typically coupled
with machine vision systems to form an automated,
precision-spray weeding machine. For precision-
spray systems to be effective, high levels of crop/
weed differentiation, accurate spray prescription
maps, knowledge of sprayer tip location relative to
target weed location, accurate herbicide placement,
and control of spray drift are needed. Although
progress and promising research has been conducted
in all of these areas, the technological challenge of
meeting all of these criteria has not been overcome
yet. An exception to this would be the case of
automated thinning machines. The following is a
brief review of some of the efforts made to date for
precision spray weeding systems and automated
thinning technology.
One of the earliest automated precision spray
weeding systems reported is the one described by
Lee et al. (1999). They developed a real-time robot
for controlling weeds in tomato crops. The device
utilized a machine vision system for detecting plants
and a sprayer capable of treating grid cell areas
measuring 0.63 by 1.25 cm. When field tested at a
speed of 0.22 m s
, the average error between the
center of the target and spray droplets delivered was
only 6.6 mm, with a standard deviation of 4.9 mm.
However, the system correctly recognized only 76%
of the tomato plants, and 52% of the weeds were
not sprayed.
Lamm et al. (2002) used the prototype machine
developed by Lee et al. (1999) as a basis for
developing a precision weeding machine for cotton.
Images were processed with the use of the excessive
green index (ExG) for plant/nonplant segmentation
and an Erosion technique for crop/weed recogni-
tion. ExG is a greenness index, calculated on a
point-by-point basis in a color image as ExG ¼2*
Green – 1 *Red, that can be used to distinguish
green plant material from nongreen backgrounds
like soil automatically. The Erosion image process-
ing technique can then iteratively remove leaf-edge
pixels to erase monocot weed leaves selectively while
retaining dicot crop leaves (e.g., cotton) for
automated crop vs. weed discrimination. When
field tested at a travel speed of 0.45 m s
, the
authors reported that 88% of the weeds were
sprayed. Further work was needed; however, as 21%
of the cotton plants were also sprayed.
Søgaard and Lund (2007) developed an autono-
mous robot for precision spraying. The system was
tested in indoor laboratory conditions with 110-m
black circular disks as targets at travel speeds of 0.2
. Study results showed that the system was
capable of autonomously delivering microdoses (2.5
ll) of spray to targets with subcentimeter accuracy.
The system was further tested in field trials planted
to oilseed rape as a test weed (Søgaard et al. 2006).
In the study, plant surface area was found to have a
large effect on machine performance. When leaf
830 Weed Technology 30, October–December 2016
area was greater than 100 mm
, targeting perfor-
mance was acceptable as more than 86% of weeds
were sprayed. When leaf area was less than 75 mm
however, targeting efficiency was less than 68%.
Overall, 82% of the weeds were effectively sprayed.
Nieuwenhuizen et al. (2010) developed a tractor-
pulled, precision weeding machine for controlling
volunteer potatoes in sugar beets. The system was
principally comprised of trigger-activated cameras
that recorded RGB and near-infrared (NIR) images,
ultrasonic sensors, a computer and a microsprayer.
Images were captured through trigger activation of
the cameras by a wheel encoder. Images were
processed by first detecting vegetation with an
excessive green threshold, and then extracting color
features for grid cells measuring 11 by 11 pixels (11
by 11 mm). A classification algorithm was used to
separate volunteer potato from sugar beet in each
grid cell and prescription maps measuring 11 by 40
mm were created. A microsprayer comprising five
needles, each connected to hoses and fast-acting
solenoid valves, was used to control weeds through
emission of 20 65ll droplets of a 5% v/v solution
of glyphosate. In laboratory tests, spray droplet
control accuracy was 614 mm in the longitudinal
direction and 67.5 mm in the transverse direction.
When field tested at travel speeds ranging from 0.2
to 0.8 m s
, the study found that on average, over
99% of sugar beet plants were left unsprayed,
whereas 86% of the volunteer potatoes plants were
sprayed. Mortality rates for sugar beet and volunteer
potato were 1% and 77%, respectively. These
results are very encouraging, however, to be useful
for most specialty crops, spray resolution, and weed
detection accuracies greater than 11 by 40 mm will
likely be needed.
Midtiby et al. (2011) developed a real-time
microspraying weed control system that utilized an
inkjet printer head for the spray assembly. The
unit’s machine vision system utilized color and
shape features to identify crop and weed plants. The
automated machine was tested with corn as the crop
plant in a laboratory setting at speeds of 0.5 m s
A spray solution containing a 5 g L
of glyphosate was used to control weeds. Although
all of the corn plants exhibited normal growth, only
37% of one of the weed species tested was
effectively controlled. Potential reasons stated for
this were suboptimal weed identification, problems
related to depositing enough herbicide on weed leaf
surfaces, or both.
Automated lettuce thinners can be thought of as
precision-spray, intrarow weeding machines, be-
cause they remove closely spaced, undesired plants
in the seed row. Currently, there are four companies
that manufacture automated thinning machines in
the United States. These include units from Ramsay
Highlander Inc. (Gonzales, CA), Agmechtronix
LLC (Silver City, NM), Blue River Technologies
Inc. (Sunnyvale, CA), and Vision Robotics Corp.
(San Diego, CA) A representative image of the
technology is shown in Figure 4. All manufacturers
use a machine vision system to locate lettuce plants
for selective thinning and herbicidal spray solutions
to kill unwanted plants. The machine vision systems
employed are designed to locate crop plants
Figure 4. Lettuce thinning at Salinas, CA with the BlueRiver
Lettuce Bot (top) and a close-up of the newly thinned lettuce
showing the sprayed zones where lettuce will be removed and
unsprayed zones where ‘‘saved’’ lettuce plants will be grown to
maturity (bottom). (Color for this figure is available in the online
version of this article.)
Fennimore et al.: Weed control automation in specialty crops 831
primarily by using color and plant size criteria
(Siemens 2014). This scheme works well, as lettuce
seedlings are typically much larger than weeds at the
time of thinning. In conditions where weed density
is high or when weeds are comparable in color or
size to that of lettuce seedlings, machine perfor-
mance is less than optimal and sometimes ineffec-
tive. To minimize spray drift, spray delivery systems
are housed in enclosures. Spray is delivered
intermittently based on desired plant spacing and
machine location through activation of fast-acting
solenoids. The location of the spray nozzle relative
to the lettuce plants is determined optometrically
with the use of an approach similar to that described
by Siemens et al. (2012), through analysis of
consecutive images or a combination of both.
Commercial lettuce thinners are typically used to
thin small plants, roughly 1.9 cm in diameter, at
speeds of 0.7 to 1.3 m s
. An example of a field
where rows of lettuce have been intermittently
sprayed by an automated thinner is shown in Figure
4. Average plant spacing can be as low as 3.8 cm,
but plant spacing greater than 4.4 cm is preferred
for optimal performance (N Abranyi, Blue River
Technologies Inc., personal communication). In
optimal conditions, targeting accuracy can be as low
as 0.6 cm, but accuracies of around 1.3 cm are
typical. Products such as fertilizer solutions, sulfuric
acid and carfentrazone are successfully used to kill
lettuce seedlings. Surfactants and/or antidrift agents
are commonly mixed with the herbicidal solution to
promote mortality of sprayed plants and prevent
damage to crop plants.
Because automated thinning technology is so
new, there are few reports in the literature regarding
their performance or economic viability. Siemens et
al. (2012) developed and evaluated a prototype
automated thinning machine in lettuce at travel
speeds of 0.7 m s
. As compared to the hand-
thinned treatment, there were no significant
differences in plant spacing, plant spacing unifor-
mity, plant population, or time required for a hand
laborer to remove plants missed during thinning.
Yields were also not significantly different between
the two treatments.
Chu et al. (2016) evaluated the performance of a
commercial automated thinning machine in ro-
maine lettuce. In the study, seeds were planted 6.3
cm apart and plants were thinned with the use of
fertilizer solutions at the two-leaf stage of growth.
Study results showed that plant size was not affected
1 wk after hand thinning and that plants were
actually larger, as compared to the hand-thinning
treatment 2 to 3 wk after hand thinning. A possible
explanation for the larger plant size with the
machine-thinned treatment was that seedling roots
were not disturbed, because plants were thinned
using chemicals and not by hand hoeing. The study
also found that automated thinning resulted in
more uniform plant spacing and increased individ-
ual plant weights. A significant yield improvement
was found at one of the three test sites.
Smith et al. (2014) evaluated the performance of
automated lettuce thinners and their impact on
labor savings. In the study, three different automat-
ed thinners were tested in commercial lettuce fields.
Like Chu et al. (2016), they also found that
automated thinning improved plant spacing uni-
formity. Although yields were moderately higher at
six of the seven test sites, automated thinners left
seven times as many closely spaced lettuce plants as
compared to the hand-thinned treatments. As a
consequence, the time required for the subsequent
hand-weeding operation, where closely spaced
plants and weeds are removed, increased by 0.5 h
. Total labor requirements for both operations,
thinning and hand weeding after thinning, however,
was 1.8 h ha
lower for the automated thinner
treatments because of higher labor use efficiency
during thinning. Assuming an hourly wage of $13
, this translates into a $23 ha
savings in labor
These studies on automated thinning machine
performance show the viability of using spray-based
systems to control unwanted plants in close
proximity to crop plants. In order for these
technologies to be used for general-purpose weed-
ing, improved machine vision systems that can
reliably differentiate between crop and weed plants
are needed. Spray systems that can accurately spot-
apply herbicides to weeds at the 1-cm scale at
commercially viable speeds also need to be
developed if weed removal accuracy is to be
Adapting Cropping Systems for Automation
Weed control technologies are commonly de-
signed and developed to function in the existing
832 Weed Technology 30, October–December 2016
characteristics of cropping systems, and this has
been the predominant approach for mechanical
weed control. Thus, equipment and operation
settings are limited by factors such as planting
density and arrangement, crop susceptibility to
mechanical injury, crop architecture, size, and
growth stage (Forcella 2012; Mohler 2001; Slaugh-
ter et al. 2008). The successful development and
adoption of automation technologies for weed
control would increase if consideration is given to
agronomic practices to find an optimum balance
between the requirements of the crop and those of
automation technologies. This two-way approach,
in fact, has been used for herbicide development.
Although herbicide registration is done predomi-
nantly for crops that are tolerant to the herbicide,
i.e., native traits, new varieties are usually released
and adopted by growers only if they have the
necessary tolerance to the herbicides commonly
used in the crop (Leon and Tillman 2015). Row
spacing in orchards and vineyards are determined
not only by canopy light interception efficiency, but
by machinery requirements for production practices
such as pruning and harvest (Wagenmakers and
Wertheim 1991). An example would be to plant
orchard crops in hedgerows to facilitate operations
such as mechanical harvesting and pruning (Connor
et al. 2014). Cross cultivation using traditional
cultivators is a very effective means of weed control
and is used in crops like perennial plantings of
artichoke (Cynara scolymus L.) (Haar et al. 2001).
The potential to modify crop planting patterns to
permit rapid cross cultivation has yet to be explored
for most crops. For example, a cultivator set up with
RTK GPS or a machine vision guidance system
would be able to cultivate a field quickly in two
directions, i.e., cross cultivation. Therefore, new
automation technologies for weed control might
need modifications in existing production practices
to take full advantage of their benefits (Rask and
Kristoffersen 2007).
Planting Arrangement. A major benefit of auto-
mation technologies is the increase in speed during
implementation of weed control, especially me-
chanical practices (Tillet et al. 2008). Although
weed suppression is higher as crop density increases
and row and intrarow spacings decrease (Boyd et al.
2009; Teasdale and Frank 1983), control practices
such as cultivation, and especially intrarow cultiva-
tion, can be greatly limited at high planting
densities and narrow spacing (Mohler 2001). Thus,
to ensure adequate intrarow weed control with the
use of mechanical devices, it might be necessary to
find an optimum planting density that does not
maximize crop weed suppression, but would
facilitate weed and crop detection and the imple-
mentation of the weed control actuators while
ensuring high levels of weed removal and crop safety
(Scarlett 2001; Slaughter et al. 2008). Another
strategy is the use of small autonomous robots that
identify and control weeds by using machine vision
to identify weeds and a weeding tool ( ˚
Astrand and
Baerveldt 2002). Autonomous vehicles that work
slowly and carefully could more accurately differ-
entiate weeds and remove them from the crop.
Studies of new automation machines should
evaluate not only different speeds, but also different
distances between crop plants.
One new piece of technology is the ‘‘plant tape’’
precision transplanter (Figure 5). The system allows
rapid and efficient transplanting of vegetable crops,
which can facilitate precision cross cultivation (Maw
and Suggs 1984; Fennimore et al. 2014). Precise
and accurate placement of the transplants allows for
easier plant tracking by intelligent cultivator
machine vision systems, which should improve
their performance.
Crop Morphology and Development. Crop
damage is perhaps the main concern for growers
and the most important limitation for the use of
mechanical weed control tools (Pleasant et al. 1994;
Vangessel et al. 1998). As explained before,
automation and robotic weed control machines will
be developed with the goal to minimize contact
with the crop. However, it is likely that this contact
will not be zero. Therefore, having cultivars that are
better suited to deal with mechanical contact by
these machines without suffering negative effects on
growth or yield will allow the implementation of
new automation technologies.
Because intrarow cultivation is necessary for
adequate weed control, especially in the absence of
herbicides such as in organic systems, the risk of
crop root damage becomes a major limitation for
this practice (Kurstjens et al. 2004; Pleasant et al.
1994). Breeding for crop plants with deeper root
systems and stronger anchorage to the soil will
greatly favor the development of mechanical
control technologies (Kurstjens et al. 2004). The
Fennimore et al.: Weed control automation in specialty crops 833
selection of germplasm with deeper root systems is
currently under way, mainly through efforts to
develop drought-tolerant varieties (Araus et al.
2002; Nibau et al. 2008). Collaborating with plant
breeders is critical to identifying germplasm that
will enable the use of automation technologies for
weed control.
Automation of weed removal appears to be a very
promising technology for specialty crop weed
management programs. The ability to spray
automatically or hoe weeds robotically, leaving the
crop undamaged, creates opportunities to increase
the efficiency of labor use in specialty crops greatly.
Labor availability for hand weeding, a critical need
in many specialty crops, is diminishing, and there is
little prospect that new herbicides will be the
remedy. Instead, advances in specialty crop weed
control can be found in the area of computer
processing, machine vision, and precise weed
destruction. Commercial weeding machines are
primarily being manufactured by European com-
panies and startup ventures in the United States
with the aim of reducing field labor costs in
specialty crops. However, there is the potential for
application of much of this machinery in agronomic
crops, especially in organic production systems.
Much of the actuator and sensor technology
presented here is not new. However, what is new
is the commercialization of weed removal technol-
ogy that integrates sprayers or intrarow cultivators
into intelligent control systems that are capable of
removal of most weeds in vegetable crops (Lati et al.
2016). It is not possible to predict the future
trajectory of commercial development of intelligent
weed removal equipment, but it is possible to say
that the economic and regulatory constraints on
herbicide development are much greater than for
intelligent weed removal technology. Companies
that are developing intelligent weed removal
technologies are relatively small compared to
traditional pesticide companies, and less constrained
by legacy herbicides that pesticide companies own
and must defend. It appears possible that small
innovative companies may be the primary source of
new weed management technology in the future.
Based on the vast improvements in robotics and
processing, it would appear that the future of
automation in weed control is very promising.
Literature Cited
Araus JL, Slafer GA, Reynolds MP, Royo C (2002) Plant
breeding and drought in C
cereals: what should we breed for?
Ann Bot 89:925–940
Astrand B, Baerveldt A (2002) An agricultural mobile robot with
vision-based perception for mechanical weed control. Auton
Robots 13:21–35
Astrand B, Baerveldt AJ (2005) Plant recognition and
localization using context information and individual plant
features. Paper IV in Vision Based Perception for Mechatronic
Weed Control. ˚
Astrand, B. PhD thesis. Chalmers University
of Technology. G¨
oteborg, Sweden.
d.lasso?id¼Fil-000788. Accessed on April 19, 2016
Boyd NB, Brennan EB, Fennimore SA (2006) Stale seedbed
techniques for organic vegetable production. Weed Technol
Figure 5. The Plant Tape machine transplanting lettuce with
six plant lines per 2-m-wide bed (top) at Spreckles, CA. Shown
on the bottom is a close up of one plant tape feeding into the
transplant mechanism. (Color for this figure is available in the
online version of this article.)
834 Weed Technology 30, October–December 2016
Boyd NS, Brennan EB, Smith RF, Yokota R (2009) Effect of
seeding rate and planting arrangement on rye cover crop and
weed growth. Agron J 101:47–51
Castello O, Gerwick C, Johnson T, LePlae PR Jr, Lo W, Roth JJ
(2015) Herbicide discovery: the search for new modes of
action. Proc Calif Weed Sci Soc.
Herbicide%20Discovery.pdf. Accessed April 19, 2016
Chu Q, Liu J, Bali K, Thorp KR, Smith R, Wang G (2016)
Automated thinning increases uniformity of in-row spacing
and plant size in romaine lettuce. HortTechnology 26:12–19
Cloutier DC, van der Weide RY, Peruzzi A, LeBlanc ML (2007)
Mechanical weed management. Pages 111–134 in Upadhyaya
MK, Blackshaw RE, eds. Non-Chemical Weed Management.
Oxfordshire, UK: CAB International
Connor DJ, G´
omez-del-Campo M, Rousseaux MC, Searles PS
(2014) Structure, management and productivity of hedgerow
olive orchards: a review. Sci Hortic 169:71–93
De Baerdemaeker J (2014) Future adoption of automation in
weed control. Chapter 13 in Young SL, Pierce FJ, eds.
Automation: The Future of Weed Control in Cropping
Systems. Dordrecht, The Netherlands: Springer Science þ
Business Media
Duke SO (2012) Why have no new herbicide modes of action
appeared in recent years? Pest Manag Sci 68:505–512
Ehsani MR, Upadhyaya SK, Mattson ML (2004) Seed location
mapping using RTK GPS. Trans ASAE 47:909–914
Fennimore SA, Doohan DJ (2008) The challenges of specialty
crop weed control, future directions. Weed Technol 22:364–
Fennimore SA, Smith RF, Tourte L, LeStrange M, Rachuy JS
(2014) Evaluation and economics of a rotating cultivator in
bok choy, celery, lettuce, and radicchio. Weed Technol
Fennimore SA, Tourte LJ, Rachuy JS, Smith RF, George CA
(2010) Evaluation and economics of a machine-vision guided
cultivation program in broccoli and lettuce. Weed Technol
Forcella F (2009) Potential of air-propelled abrasives for selective
weed control. Weed Technol 23:317–320
Forcella F (2012) Air-propelled abrasive grit for postemergence
in-row weed control in field corn. Weed Technol 26:161–164
Forcella F (2013) Soybean seedlings tolerate abrasion from air-
propelled grit. Weed Technol 27:631–635
Garrett RE, Talley WK (1969) Use of gamma ray transmission
in selecting lettuce for harvest, Paper No. 69-310, The 1969
Annual Meeting of the American Society of Agricultural
Gianessi LP, Reigner NP (2007) The value of herbicides in U.S.
crop production. Weed Technol 21:559–566
Haar MJ, Fennimore SA (2003) Evaluation of integrated
practices for common purslane management in lettuce. Weed
Technol 17:229–233
Haar MJ, Fennimore SA, Lambert CL (2001) Economics of
pronamide and pendimethalin use in artichoke during stand
establishment. HortScience 36:650–653
Haff RP, Slaughter DC, Jackson ES (2011) X-ray based stem
detection in an automatic tomato weeding system. Appl Eng
Agric 27:803–810
Have H, Jon Nielsen J, Blackmore S, Theilby F (2005)
Autonomous Weeder for Christmas Trees —Basic Develop-
ment and Tests. Pesticides Research No. 97 2005 Bekæmpel-
sesm iddelforskning fra Miljøstyrelsen. Danish Environmental
Protection Agency.
869-6.pdf. Accessed April 28, 2016. 77 pp
Hearn DJ (2009) Shape analysis for the automated identification
of plants from images of leaves. Taxon 58:934–954
Hemming J, Nieuwenhuizen AT, Struik LE (2011) Image
analysis system to determine crop row and plant positions for
an intra-row weeding machine. Tokyo, Japan: Proceedings of
the CIGR International Symposium on Sustainable Biopro-
duction. Pp 1–7
Jeschke P (2016) Progress of modern agricultural chemistry and
future prospects. Pest Manage Sci 72:433–455
Kaierle S, Marx C, Rath T, Hustedt M (2013) Find and
irradiate—lasers used for weed control. Laser Tech J 10:44–47
Knezevic S, Fennimore S, Datta A (in press) Thermal weed
control. In Encyclopedia of Applied Plant Sciences. 2nd edn.
Kraehmer H, Laber B, Rosinger C, Schulz A (2014) Herbicides
as weed control agents: state of the art: I. Weed control
research and safener technology: the path to modern
agriculture. Plant Physiol 166:1119–1131
Kurstjens DAG, Kropff MJ, Perdok UD (2004) Method for
predicting selective uprooting by mechanical weeders from
plant anchorage forces. Weed Sci 52:123–132
Lamberth C, Jeanmart S, Luksch, Plant A (2013) Current
challenges and trends in the discovery of agrochemicals.
Science 341:742–746
Lamm RD, Slaughter DC, Giles DK (2002) Precision weed
control system for cotton. Trans ASAE 45:231–238
Lanini WT, LeStrange M (1991) Low-input management of
weeds in vegetable fields. Calif Agric 45:11–13
Lati RN, Siemens MC, Rachuy JS, Fennimore SA (2016) Intra-
row weed removal in broccoli and transplanted lettuce with an
intelligent cultivator. Weed Technol 30:655–663
Lee WS, Slaughter DC, Giles DK (1999) Robotic weed control
system for tomatoes. Precis Agric 1:95–113
Lenker DH, Adrian PA (1971) Use of x-rays for selecting mature
lettuce heads. Trans ASAE 14:894–898
Leon RG, Tillman BL (2015) Postemergence herbicide tolerance
variation in peanut germplasm. Weed Sci 63:546–554
Manh AG, Rabatel G, Assemat L, Aldon MJ (2001) Weed leaf
image segmentation by deformable templates. J Agric Eng Res
Mathiassen SK, Bak T, Christensen S, Kudsk P (2006) The
effect of laser treatment as a weed control method. Biosyst Eng
Maw BW, Suggs CW (1984) A seedling taping machine for bare
root plants. Trans ASAE 27:711–714
Melander B, Lattanzi B, Pannaci E (2105) Intelligent versus non-
intelligent mechanical intra-row weed control in transplanted
onion and cabbage. Crop Prot 72:1–8
Melita CD, Muscato G, Poncelet M (2013) A simulation
environment for an augmented global navigation satellite
system assisted autonomous robotic lawn-mower. J Intell
Robot Syst 71:127–142
Fennimore et al.: Weed control automation in specialty crops 835
Michaels A, Haug S, Albert A (2015) Vision-based high-speed
manipulation for robotic ultra-precise weed control. Pages
5498–5505 in Intelligent Robots and Systems (IROS), 2015
IEEE/RSJ International Conference.
Midtiby HS, Mathiassen SK, Andersson KJ Jørgensen RN
(2011) Performance evaluation of a crop/weed discriminating
microsprayer. Comput Electron Agric 77:35–40
Mohler CL (2001) Mechanical management of weeds. Pages
139–209 in Liebman M, Mohler CL, Staver CP, eds.
Ecological management of agricultural weeds. Cambridge,
United Kingdom: Cambridge University Press
Nibau C, Gibbs DJ, Coates JC (2008) Branching out in new
directions: the control of root architecture by lateral root
formation. New Phytol 179:595–614
Performance evaluation of an automated detection and control
system for volunteer potatoes in sugar fields. Biosyst Eng
Nørremark M, Griepentrog HW, Nielsen J, Søgaard HT (2008)
The development and assessment of the accuracy of an
autonomous GPS-based system for intra-row mechanical weed
control in row crops. Biosyst Eng 101:396–410
Nørremark M, Søgaard HT, Griepentrog HW, Nielsen H
(2007) Instrumentation and method for high accuracy
georeferencing of sugar beet plants. Comput Electron Agric
Nørremark M, Swain KS, Melander B (2009) Advanced non-
chemical and close to plant weed control system for organic
agriculture. In International Agricultural Engineering Confer-
ence, Bangkok.
O’Dogherty MJ, Godwin RJ, Dedousis AP, Brighton LJ, Tillett
ND (2007) A mathematical model of the kinematics of a
rotating disc for inter- and intra-row hoeing. Biosyst Eng
Onyango CM, Marchant JA (2003) Segmentation of row crop
plants from weeds using colour and morphology. Comput
Electron Agric 39:141–155
erez-Ruiz M, Slaughter DC, Gliever CJ, Upadhyaya SK
(2012a) Automatic GPS-based intra-row weed knife control
system for transplanted row crops. Comput Electron Agric
erez-Ruiz M, Slaughter DC, Gliever C, Upadhyaya SK (2012b)
Tractor-based real-time kinematic–global positioning system
(RTK-GPS) guidance system for geospatial mapping of row
crop transplant. Biosyst Eng 111:64–71
Persson M, ˚
Astrand B (2008) Classification of crops and weeds
extracted by active shape models. Biosyst Eng 100:484–497.
DOI: 10.1016/j.biosystemseng.2008.05.003
Pleasant JM, Burt RF, Frisch JC (1994) Integrating mechanical
and chemical weed management in corn (Zea mays). Weed
Technol 8:217–223
Poulsen, F. (2011) System for selective treatment of plants in a
row. U.S. patent 8,027,770 B2
Rask AM, Kristoffersen P (2007) A review of non-chemical weed
control on hard surfaces. Weed Res 47:370–380
uegg WT, Quadranti M, Zoschke A (2007) Herbicide research
and development: challenges and opportunities. Weed Res
Scarlett AJ (2001) Integrated control of agricultural tractors and
implements: a review of potential opportunities relating to
cultivation and crop establishment machinery. Comput
Electron Agric 30:167–191
Shaner DL (2000) The impact of glyphosate-tolerant crops on
the use of other herbicides and on resistance management. Pest
Manage Sci 56:320–326
Siemens MC (2014) Robotic weed control. Pages 76–80 in
Proceedings of the California Weed Science Society. Vol. 66.
Salinas, CA: California Weed Science Society.
Siemens MC, Herbon R, Gayler RR, Nolte KD, Brooks D
(2012) Automated machine for thinning lettuce—evaluation
and development. ASABE paper No. 12-1338169. St. Joseph,
Slaughter DC, Chen P, Curley RG (1999) Vision guided
precision cultivation. Precis Agric 1:199–216
Slaughter DC, Giles DK, Fennimore SA, Smith RF (2008)
Multispectral machine vision identification of lettuce and weed
seedlings for automated weed control. Weed Technol 22:378–
Smith R, Mosqueda, E, Shrestha A, Galva, F (2014) Evaluation
of automated lettuce thinners. Salinas Valley Agriculture–
ANR Blogs. Davis, CA: University of California, Davis. http://¼15932.
Accessed April 28, 2016
Søgaard HT (2005) Weed classification by active shape models.
Biosyst Eng 91:271–281
Søgaard HT, Lund I (2007) Application accuracy of a machine
vision-controlled robotic micro-dosing system. Biosyst Eng
Søgaard HT, Lund I, Graglia E (2006) Real-time application of
herbicides in seed lines by computer vision and micro-spray
system. ASABE paper No. 06-1118, pp. 9. St. Joseph, MI:
Southall B, Hague T, Marchant JA, Buxton BF (2002) An
autonomous crop treatment robot: Part I. A Kalman filter
model for localization and crop/weed classification. Int J
Robot Res 21:61–74
Sun H, Slaughter DC, Perez-Ruiz M, Gliever C, Upadhyaya SK,
Smith RF (2010) RTK GPS mapping of transplanted row
crops. Comput Electron Agric 71:32–37
Taylor JE, Charlton D, Yunez-Naude A (2012) The end of farm
labor abundance. Appl Econ Perspect Policy 34:587–598
Teasdale JR, Frank JR (1983) Effect of row spacing on weed
competition with snap beans (Phaseolus vulgaris). Weed Sci
Tillet ND, Hague T, Grundy AC, Dedousis AP (2008)
Mechanical within-row weed control for transplanted crops
using computer vision. Biosyst Eng 99:171–178
Tillett ND, Hague T, Miles SJ (2001) A field assessment of a
potential method for weed and crop mapping on the basis of
crop planting geometry. Comput Electron Agric 32:229–246
Upadhyaya SK, Ehsani M, Mattson ML (2003) Method and
apparatus for ultra precise GPS-based mapping of seeds or
vegetation during planting. U.S. Patent 6,553,312
Upadhyaya SK, Ehsani M, Mattson ML (2005) Method and
apparatus for ultra precise GPS-based mapping of seeds or
vegetation during planting. U.S. patent 6941225
836 Weed Technology 30, October–December 2016
[USDA] U.S. Department of Agriculture (2016) What is a
Specialty Crop?
scbgp/specialty-crop. Accessed April 28, 2016
[USEPA] U.S. Environmental Protection Agency (2015)
Pesticide Registration Manual: Chapter 13—Devices. http://
manual-chapter-13-devices. Accessed January 23, 2016
[USGS] U.S. Geological Survey (2016) Global Positioning
Application and Practice. Ac-
cessed March 25, 2016
van der Schans D, Bleeker P, Molendijk L, Plentinger M, van der
Weide R, Lotz B, Bauermeister R, Total R, Baumann DT
(2006) Practical Weed Control in Arable Farming and
Outdoor Vegetable Cultivation without Chemicals. PPO
Publication 532. Applied Plant Research, Wageningen
University, Lelystad, the Netherlands. 77 pp
Vangessel MJ, Schweizer EE, Wilson RG, Wiles LJ, Westra P
(1998) Impact of timing and frequency of in-row cultivation
for weed control in dry bean (Phaseolus vulgaris). Weed
Technol 12:548–553
Wagenmakers PS, Wertheim SJ (1991) Planting systems for fruit
trees in temperate climates. Circ Rev Plant Sci 10:369–385
Wortman S (2014) Integrating weed and vegetable crop
management with multifunctional air-propelled abrasive grits.
Weed Technol 28:243–252
Wortman S (2015) Air-propelled abrasive grits reduce weed
abundance and increase yields in organic vegetable production.
Crop Prot 77:157–162
Zhang Y, Slaughter DC (2011) Hyperspectral species mapping
for automatic weed control in tomato under thermal
environmental stress. Comput Electron Agric 77:95–104.
DOI: 10.1016/j.compag.2011.04.001
Zhang Y, Slaughter DC, Staab ES (2012a) Robust hyperspectral
vision-based classification for multi-season weed mapping. J
Photogramm Remote Sens 69:65–73. DOI: 10.1016/j.
Zhang Y, Staab ES, Slaughter DC, Giles DK, Downey D
(2012b) Automated weed control in organic row crops using
hyperspectral species identification and thermal micro-dosing.
Crop Prot 41:96–105
Received April 28, 2016, and approved July 4, 2016.
Associate editor for this paper: Robert Nurse, Agriculture
and Agri-Food Canada.
Fennimore et al.: Weed control automation in specialty crops 837
... Currently, there are two main weeding methods, namely, mechanical weeding and chemical weeding. Mechanical weeding has poor adaptability to topography and agronomy [2,3], high energy consumption [4], and low operation efficiency [5], and some weed types cannot be completely eradicated by mechanical weeding [6]. Chemical weeding is the most widely used weeding method by farmers worldwide [7][8][9]. ...
Full-text available
To address the problem of herbicide residues exceeding the safety standard due to continuous spraying of herbicides on open-field cabbage, we propose an intermittent weed spraying control method integrating cabbage position, cabbage canopy size, and spraying machine operation speed. It is based on an early-stage cabbage target identification method obtained in the early stage and the operation requirements in open-field cabbage. Built with a C37 controller, a stable pressure spray system and an intermittent weed spraying control system for open-field cabbage, an integrated system was designed. Experimental verification was carried out through measurement indexes such as spraying precision, herbicide saving rate, herbicide efficacy, and herbicide residue. Since the industry is faced with a status quo of a lack of relevant operational norms and national standards for the precise weed spraying operation mode, this paper provides a relatively perfect experiment and evaluation method for this mode. The experimental results on the accuracy of weed spraying at different speeds showed that the mean absolute error (MAE), root mean square error (RMSE), and average spray cabbage coverage rate (ASCCR) of intermittent weed spraying increased, but the average effective spray coverage rate (AESCR) decreased with increasing operation speed. When the working speed was 0.51 m/s, the MAE and RMSE of intermittent weed spraying were less than 2.87 cm and 3.40 cm, respectively, and the AESCR was 98.4%, which verified the feasibility of operating the intermittent weed spraying of cabbage. The results of a field experiment showed that the average weed-killing rate of intermittent weed spraying for open-field cabbage was 94.8%, and the herbicide-saving rate could reach 28.3% for a similar weeding effect to that of constant-rate application, which not only met the needs of intermittent weed spraying in open-field cabbage but also had great significance for improving the herbicide utilization rate. Compared with the constant-rate application method, the herbicide residue concentration detected using intermittent weed spraying for cabbage decreased by 66.6% on average, which has important research significance and application value for ensuring the normal growth of crops and the safety of agricultural products.
... Weeds can be controlled more effectively with hoeing than with raking, and hoe knives are better at tackling taproot weeds. Smart cultivator systems that employ GPS or precision sensor cameras can be used to guide the hoe blades so they can get near to the crops without endangering them (Fennimore et al., 2016;Machleb et al., 2020). ...
Conference Paper
Full-text available
Weeds, one of the most significant pests of agricultural products, substantially impair product quality while also reducing crop productivity. If preventive weed management techniques fail to effectively control weeds in the field, direct weed control may be required. Although herbicides are the most often used direct weed control method, farmers can also use non-chemical weed control methods. Non-chemical weed management techniques are now more important than ever due to several variables. Among these are issues with the environment brought on by the excessive use of herbicides, the rise in the demand for organic food, and the development of herbicide resistance in weeds. Herbicide use and risk can be reduced by conventional farmers by using mechanical and biological control methods that are frequently utilized on organic farms. Consequently, cultural treatments, mechanical methods, and biological tactics are examples of non-chemical weed control measures. Recent years have seen the rise of robotic technology, cover crops, and thermal procedures among these methods. This paper explains why non-chemical weed control is necessary. This offers a summary of the current non-chemical weed control methods that are available.
... For example, lightweight robots can be used in the field when it is too wet for conventional equipment, allowing for more timely weed control. These systems can also dramatically reduce the overall cost of weed control, particularly in organic production or in specialty crops where there are fewer herbicide options (Fennimore et al. 2017;Lowenberg-DeBoer et al. 2020). ...
Full-text available
Recent innovations in 3-D imaging technology have created unprecedented potential for better understanding weed responses to management tactics. Although traditional 2-D imaging methods for mapping weed populations can be limited in the field by factors such as shadows and tissue overlap, 3-D imaging mitigates these challenges by using depth data to create accurate plant models. Three-dimensional imaging can be used to generate spatio-temporal maps of weed populations in the field and target weeds for site-specific weed management, including automated precision weed control. This technology will also help growers monitor cover crop performance for weed suppression and detect late-season weed escapes for timely control, thereby reducing seed bank persistence and slowing the evolution of herbicide resistance. In addition to its many applications in weed management, 3-D imaging offers weed researchers new tools for understanding spatial and temporal heterogeneity in weed responses to IWM tactics, including weed-crop competition and weed community dynamics. This technology will provide simple and low-cost tools for growers and researchers alike to better understand weed responses in diverse agronomic contexts, which will aid in reducing herbicide use, mitigating herbicide resistance evolution, and improving environmental health.
Advancements of the last decade in edge computing, edge IoT, and edge artificial intelligence now allow for autonomous, efficient, and intelligent systems to be proposed for various industrial applications. Intelligence agricultural solutions allow farmers to achieve more with less while improving quality and providing a rapid go-to-market approach for produce. Using AI is an effective technique to detect any crop health concerns or nutrient inadequacies in the field. Plant diseases affect the food system, economy, and environment. This chapter covers intelligent agriculture & challenges in front of technology. It focuses AI application using machine learning, artificial neural network (ANN), and deep learning. The various AI applications in agriculture for land monitoring, crop and varietal selection, smart irrigation or automation of irrigation, monitoring of crop health, crop disease detection, predictive analytics, weed control, precision agriculture, harvesting, yield estimation and phenotyping, supply chain management, and food quality.
Agriculture sustainability is among the core elements of the fourth industrial revolution (4IR) in line to meet the challenges of increasing demands of the rising global population (expected to be 9.9 billion by the year 2050) and climate change. Development and adaptation of innovative technologies can be perceived as a dynamic approach that can help in enhancing agricultural outputs and mitigating the threat of food shortages worldwide. For this, integration of advanced technologies such as precision agriculture, artificial intelligence (AI), the Internet of Things (IoT), machine learning, big data analytics, cloud computing, drone tech, OMICs, and genome editing technologies, especially the CRISPR-CAS9-based genetic engineering, is turning into a pivotal tool for sustainable production, crop management, quality control, integrated disease management, and overall development of agricultural sector. Moreover, digital transformations have also paved the path of sustainability-driven agricultural challenges and their solutions. In the same way, the rise of emerging technological innovations at the agricultural farms with the latest equipment helps in taking precise decisions for increasing crop production. Therefore, all of the above groundbreaking technologies illustrate a key step forward for precise, smart, and sustainable agriculture keeping in view the future challenges.
Suzbijanje korova jedan je od najvažnijih čimbenika u svim poljoprivrednim biljnim proizvodnjama, a osobito u ratarskoj proizvodnji zbog velikih površina na kojima se ona odvija. Nakon otkrića visoko učinkovitih herbicida suzbijanje korova u poljoprivredi godinama se provodilo gotovo isključivo kemijskim metodama, ali intenzivna uporaba kemijskih sredstava je rezultirala značajnim negativnim učincima na okoliš i ljudsko zdravlje. Veliki napredak u elektronici i računalnim tehnologijama te razvoj sustava za navođenje vozila zajedno s uvođenjem precizne poljoprivrede su otvorili mogućnost korištenja robota u suzbijanju korova. Posljednjih godina su razvijeni robotski sustavi s različitim metodama suzbijanja korova kao što su selektivna kemijska aplikacija, mehaničko uklanjanje korova, korištenje plamena, pare, električnog pražnjenja i lasera. U radu su navedeni primjeri različitih izvedbi robota za suzbijanje korova.
There are many species of Artocarpus fruits in Malaysia, which have different market potentials. This study classifies 4 species of Artocarpus fruits using deep learning approach, which is Convolutional Neural Network (CNN). A new proposed CNN model is compared with pre-trained models, i.e., VGG-16, ResNet50, and Xception. Effects of variables, i.e., hidden layers, perceptrons, filter number, optimizers, and learning rate, on the proposed model are also investigated in this study. The best performing model in this study is the new proposed model with 2 CNN layers (12, 96 filters) and 6 dense layers with 147 perceptrons, achieving an accuracy of 87%.KeywordsDeep learningTransfer learningConvolutional neural networkFruit classification Artocarpus
Machine vision‐based weed detection relies on features such as plant colour, leaf texture, shape, and patterns. Drought stress in plants can alter leaf colour and morphological features, which may in turn affect the reliability of machine vision‐based weed detection. The objective of this research was to evaluate the feasibility of using deep convolutional neural networks for the detection of Florida pusley (Richardia scabra L.) growing in drought stressed and unstressed bahiagrass (Paspalum natatum Flugge). The object detection neural networks you only look once (YOLO)v3, faster region‐based convolutional network (Faster R‐CNN), and variable filter net (VFNet) failed to effectively detect Florida pusley growing in drought stressed or unstressed bahiagrass, with F1 scores ≤0.54 in the testing dataset. Nevertheless, the use of the image classification neural networks AlexNet, GoogLeNet, and Visual Geometry Group‐Network (VGGNet) was highly effective and achieved high (≥0.97) F1 scores and recall values (≥0.98) in detecting images containing Florida pusley growing in drought stressed or unstressed bahiagrass. Overall, these results demonstrated the effectiveness of using an image classification convolutional neural network for detecting Florida pusley in drought stressed or unstressed bahiagrass. These findings illustrate the broad applicability of these neural networks for weed detection.
Full-text available
Thermal weed control is the use of heat as a method of vegetation control. Various levels of heat can either damage plant tissue or kill a plant. The critical temperature for leaf mortality ranges from 55. °C to 70. °C. The mechanism by which heat injury affects the plant varies from species to species and is complex. In general, plants die from the loss of membrane semipermeability and cuticle breakdown (resulting in plant desiccation), denaturation of proteins, water boiling, and other chemical decomposition. Thermal tools for weed control may include flaming, infrared and ultraviolet-B radiation, hot water, microwave, and steam. The general principle behind these thermal weed control methods is to cause plant death by raising temperatures high enough to boil water inside the plant tissue and disintegrate the plants' cellular membranes.
Full-text available
The low availability and high cost of farm hand labor make automated thinners a faster and cheaper alternative to hand thinning in lettuce (Lactuca sativa). However, the effects of this new technology on the uniformity of plant spacing and size as well as crop yield are not proven. Three experiments were conducted in commercial romaine heart lettuce fields in 2013 and 2014 in Imperial Valley, CA, to compare the effects of automated thinning and hand thinning on uniformity of inrow spacing, plant size, and crop yield. Overhead images taken at 1 week after hand thinning indicate that thinning 8 to 11 days earlier by automated thinners did not affect plant size compared with the hand thinning treatment. However, lettuce plants in the automated thinning treatment were larger than plants in the hand thinning treatment 2 to 3 weeks after hand thinning. Automated thinners increased the uniformity of in-row spacing, increased the percentage of plants with the desired in-row spacing of 24 to 32 cm, and almost completely removed plants with an undesirable in-row spacing of 4 to 20 cm. As a result, individual lettuce plant weight and heart weight from the automated thinning plots was significantly greater and plants were more uniform compared with the hand thinned plants. Despite increases in lettuce plant size and uniformity in all three experiments, yield benefits of automated thinning were only significant in one of the three experiments due to larger plant populations resulting from hand thinning.This study suggests that automated thinning increases lettuce plant size and uniformity and makes it possible for growers to increase plant population and crop yield by optimizing in-row spacing. © 2016, American Society for Horticultural Science. All rights reserved.
Concerns over environmental and human health impacts of conventional weed management practices, herbicide resistance in weeds, and rising costs of crop production and protection have led agricultural producers and scientists in many countries to seek strategies that take greater advantage of ecological processes and thereby allow a reduction in herbicide use. This book provides principles and practices for ecologically based weed management in a wide range of temperate and tropical farming systems. After examining weed life histories and processes determining the assembly of weed communities, the authors describe how tillage and cultivation practices, manipulations of soil conditions, competitive cultivars, crop diversification, grazing livestock, arthropod and microbial biocontrol agents, and other factors can be used to reduce weed germination, growth, competitive ability, reproduction and dispersal. Special attention is given to the evolutionary challenges that weeds pose and the roles that farmers can play in the development of new weed-management strategies.
Technical Report
Control of weeds in Christmas tree plantations is essential for good growth and quality of the trees. Weeds are usually controlled by frequent spraying, which may have undesired environmental effects. Recent mechanical weed control methods have been developed, but these must be carried out with special machines, which are expensive to use, especially because of the high labour requirement. The purpose of the present project was to develop and test a small, driverless weeding machine in order to alleviate these problems and make Christmas tree production more environmentally and economically sustainable. The starting point for development of the machinery concept was field experiments, which showed that it with regards to growth was sufficient to control weeds within a circle of 40 cm radius around each tree. This corresponds to about 40 % of the total area. Plants growing on the rest of the area may be left as they provide some shelter and may have a positive effect on the trees.
Glyphosate can now be used for selective, post-emergence weed control ol in glyphosate-tolerant varieties of soybeans, cotton, canola and maize. It is estimated that glyphosate-tolerant soybeans in the US will account for 50-80% of the area planted by 2001. The rapid acceptance of this new technology is due to multiple factors including broad-spectrum weed control, low cost: and simplicity. The use of glyphosate has resulted in a major reduction in the use of other herbicides including the ACCase inhibitors, ALS inhibitors, and Protox inhibitors, In the short term (three to five years) this change in herbicide use patterns will continue, In the long term (five to eight years), the primary reliance on glyphosate for weed control particularly in continuous cropping or in rotations of glyphosate-tolerant crops will result in a shift in the weed spectrum toward more tolerant weed species. As a result of this shift, other herbicides will be needed to fill these weed gaps, Continuous use of glyphosate may also lead to the selection of glyphosate-resistant weed populations, as has already occurred in Australia. However, shies in the weed species' composition from highly susceptible toward more tolerant species will happen more rapidly than selection of resistance, New herbicides developed in the future will have to be extremely cost-effective to compete against glyphosate and may be geared towards controlling weeds tolerant to glyphosate, There will also be further development of new tolerant crops to other broad-spectrum, non-selective herbicides that will be able to compete direct-ly with glyphosate. (C) 2000 Society of Chemical Industry.
The performance of the Robovator (F. Poulsen Engineering ApS, Hvalsø, Denmark), a commercial robotic intrarow cultivator, was evaluated in direct-seeded broccoli and transplanted lettuce during 2014 and 2015 in Salinas, CA, and Yuma, AZ. The main objective was to evaluate the crop stand after cultivation, crop yield, and weed control efficacy of the Robovator compared with a standard cultivator. A second objective was to compare hand weeding time after cultivation within a complete integrated weed management (IWM) system. Herbicides were included as a component of the IWM system. The Robovator did not reduce crop stand or marketable yield compared with the standard cultivator. The Robovator removed 18 to 41% more weeds at moderate to high weed densities and reduced hand-weeding times by 20 to 45% compared with the standard cultivator. At low weed densities there was little difference between the cultivators in terms of weed control and hand-weeding times. The lower-hand weeding time with the Robovator treatments suggest that robotic intrarow cultivators can reduce dependency on hand weeding compared with standard cultivators. Technological advancements and price reductions of these types of machines will likely improve their weed removal efficacy and the long-term viability of IWM programs that will use them. Can be accessed at -
Experiments conducted over three years compared weed cover and grain yields in corn which received cultivation alone, herbicide alone, and treatments combining mechanical and chemical weed control. Weed cover averaged 30% with cultivation alone compared to 9% for the other treatments. Grain yields were 7% lower in one year. In the cultivation-alone treatments the rolling cultivator was less effective in controlling weeds than the shovel/sweep cultivator. In-row weed cover was greater than between-row weed cover with cultivation alone. With banded herbicide plus cultivation, in and between-row weed cover was the same. Weed cover and grain yields following banded herbicide plus cultivation were equivalent to broadcast herbicide with or without cultivation.