Esposito et al. Chem. Biol. Technol. Agric.

(2021) 8:18
https://doi.org/10.1186/s40538-021-00217-8

REVIEW Open Access

Drone and sensor technology for sustainable

weed management: a review
Marco Esposito†, Mariano Crimaldi†, Valerio Cirillo* , Fabrizio Sarghini and Albino Maggio

Abstract
Weeds are amongst the most impacting abiotic factors in agriculture, causing important yield loss worldwide. Inte-
grated Weed Management coupled with the use of Unmanned Aerial Vehicles (drones), allows for Site-Specific Weed
Management, which is a highly efficient methodology as well as beneficial to the environment. The identification of
weed patches in a cultivated field can be achieved by combining image acquisition by drones and further process-
ing by machine learning techniques. Specific algorithms can be trained to manage weeds removal by Autonomous
Weeding Robot systems via herbicide spray or mechanical procedures. However, scientific and technical under-
standing of the specific goals and available technology is necessary to rapidly advance in this field. In this review, we
provide an overview of precision weed control with a focus on the potential and practical use of the most advanced
sensors available in the market. Much effort is needed to fully understand weed population dynamics and their com-
petition with crops so as to implement this approach in real agricultural contexts.
Keywords: Site-specific weed management, UAVs, Precision agriculture, Weed detection, Crop–weed interaction

Introduction mildew in crops such as wheat (Triticum aestivum L.),

Biotic threats such as insects, weeds, fungi, viruses, and oats, and barley (Hordeum vulgare L.) [14]; altamisa (Par-
bacteria can broadly affect crop yield and quality. Among thenium hysterophorus L.) can be a secondary host of the
these, weeds are the most impacting problem causing common hairy caterpillar (Diacrisia obliqua Walk.) [15,
remarkable yield loss worldwide [1]. The most char- 16]; Cyperus rotundus can host the root-knot Meloido-
acterized effect of weeds is competition for resources gyne graminicola and, therefore, can contribute to their
such as light [2], water [3], space [4], and nutrients [5]. spreading in the field [17]. Finally, weed infestation may
In addition, specific chemical signals and/or toxic mol- affect fresh and processed products quality such as beer,
ecules produced by weeds may interfere with a normal wine, forage [18, 19]. In this respect, weed residuals may
crop development [6]. A distinctive trait of wild species, cause accumulation of off-flavors products [20, 21], or in
including weeds, is their high physiological, morphologi- some cases, can make them harmful to humans and ani-
cal, and anatomical plasticity which makes them more mals [22, 23]. Weeds may also contain high levels of aller-
tolerant than crop species to environmental stressors [7– gens and/or toxic metabolites that, if ingested, can cause
10]. Moreover, weeds interact with other biological com- asthma, skin rash, and other reactions [24, 25].
ponents of the environment, acting as refuge for plant Most weed research aims at developing strategies that
pests such as insects, fungi, and bacteria that can harm can reduce the deleterious impact of the interspecific
close in crops [11–13]. For example, wild oats (Avena competition between crops and weeds and recent tech-
fatua L.) can harbor the etiological agents of the powdery nological advances may further contribute to this scope,
while improving the sustainability of weed control [26–
28]. Worldwide, weed competition causes severe yield
*Correspondence: valerio.cirillo@unina.it reduction in all major crops, such as wheat (23%), soy-
†
Marco Esposito and Mariano Crimaldi contributed equally to this work
Department of Agricultural Sciences, University of Naples Federico II, Via bean (37%), rice (37%), maize (40%), cotton (36%), and
Università 100, 80055 Portici, NA, Italy potato (30%) [1]. Yearly, weeds cause 50% yield losses of

© The Author(s) 2021. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and
the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material
in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material
is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the
permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat​iveco​
mmons​.org/licen​ses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creat​iveco​mmons​.org/publi​cdoma​in/
zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Esposito et al. Chem. Biol. Technol. Agric. (2021) 8:18 Page 2 of 11

corn and soybean productivity in North America. For integrates several agro-ecological aspects such as the role
corn, this equates to a loss of 148 million tons for an of conservation tillage and crop rotation on weeds seed
economical loss of over $26.7 billion [29]. In Australia, bank dynamics [10], the ability to forecast the critical
yield loss due to weeds accounts for 2.76 million tons of period of weed interference and their competition with
grain from different plants, including wheat, barley, oats, crops [45, 46], and the specific critical levels of crops/
canola, sorghum, and pulses [30]. The annual global eco- weeds interaction [47]. Therefore, an effective IWM must
nomic loss caused by weeds has been estimated to be rely on a thorough knowledge of crop-weeds competition
more than $100 billion U.S. dollars [31], despite world- dynamics, which currently represents one of the most
wide annual herbicide sales in the range of $25 billion active research areas in weed science [48, 49].
[32]. In Europe, herbicides are the second most-sold pes-
ticides. They accounted for 35% of all pesticide sales in New technologies for site‑specific weed management
2018, overcoming insecticides and acaricides (Fig. 1) [33]. Precision agriculture relies on technologies that combine
sensors, information systems, and informed management
Weed management requires an integrated approach to optimize crop productivity and to reduce the environ-
In 2050, the world population will quadruplicate, reach- mental impact [50]. Nowadays, precision agriculture has
ing 9.15 billion people [34]. However, the predicted a broad range of applications and it is employed in dif-
increase in food demand will be hardly met by the cur- ferent agricultural contexts including pests control [51],
rent production system [35]. Also, climate change will be fertilization, irrigation [52, 53], sowing [54] and harvest-
an additional challenge for the human food supply in the ing [55]. Precision agriculture can be effectively applied
near future [36]. Among all the processes affecting crop to IWM also. In the last decade, precision agriculture
productivity, weed management will be one of the hard- has rapidly advanced because of technological innova-
est challenges [37]. Mechanical and chemical weed con- tions in the areas of sensors [56], computer hardware
trol has disadvantages that probably will impede them [57], nanotechnology [58], unmanned vehicles systems
to be effective for future weed management [38–43]. and robots [59] that may allow for specific identification
Mechanical methods are scarcely efficient, and herbi- of weeds that are present in the field [47]. Unmanned
cides have a high ecological impact. An approach that aerial vehicles (UAV) are one of the most successful tech-
minimizes the drawbacks of mechanical and chemical nologies applied in precision agriculture [60]. Unmanned
weed control is Integrated Weed Management (IWM). Vehicles systems are mobile Aerial (UAV) or Terrestrial
IWM combines chemical, biological, mechanical, and/ (UTV) platforms that provide numerous advantages for
or crop management methods, and represents a model the execution and monitoring of farming activities [61].
to improve the efficiency and sustainability of weed con- UAVs can be highly valuable since they allow for Site-
trol [3, 44]. In contrast to traditional methods, IWM Specific Weed Management (SSWM) (Fig. 2). SSWM is

Fig. 1 Percentage (of total volume in kilograms) of pesticide sales by category in Europe in 2018 [33]
Esposito et al. Chem. Biol. Technol. Agric. (2021) 8:18 Page 3 of 11

an improved weed management approach for highly effi- resolution that other systems do not [70, 71]. From an
cient and environmentally safe control of weed popula- economic point of view, the use of drones requires the
tions [28], enabling precise and continuous monitoring investment to buy a UAV system with at least a 0.1 cm/px
and mapping of weed infestation. SSWM consents to resolution RGB camera, a trained pilot for flight manage-
optimize weed treatments for each specific agronomical ment and post-processing software capabilities. The ini-
situation [62]. The combination of UAVs with advanced tial UAV investment is compensated by the repeatability
cameras and sensors, able to discern specific weeds [63], of flights, which increases the frequency of datasets deliv-
and GPS technologies, that provide geographical infor- ered, and the higher resolution compared to other sys-
mation for field mapping, can help in precisely monitor- tems [72, 73]. UAVs systems also have further advantages:
ing large areas in a few minutes. Thanks to more accurate (1) the possibility to collect easily deployable data in real
planning of weed management that can increase mechan- time (excluding post-processing); (2) they can be used to
ical methods effectiveness and/or reduce herbicide survey areas with high level of hazard and/or difficult to
spread [64], the potential agro-ecological and economic reach; (3) they allow operators to collect data even with
implications of SSWM are remarkable, yielding lower unfavorable weather conditions, such as in very cloudy or
production costs, reducing the onset of weed resistance, foggy days, under which satellite detection systems fail or
improving biodiversity, and containing environmental produce very altered datasets [71]. The most important
impacts [65]. The application of UAVs to weed control sensors available as payload are mainly categorized into
can, therefore, contribute to improve the sustainability of three classes depending on the spectral length and num-
future agricultural production systems that must comply ber they can record:
with an increasing world population [34, 35].
• RGB (Red, Green, Blue) or VIS (Visible) sensors
UAVs remote sensing techniques and sensors • Multispectral sensors
UAVs have become a common tool in precision agricul- • Hyperspectral sensors
ture [66, 67]. Thanks to their affordability, user-friendli-
ness and versatility, UAVs are often the primary choice
for fast and precise in situ remote sensing or survey RGB/VIS sensors
operations. Despite their versatility, these systems may The RGB or VIS sensors are the most common and
be used for different purposes, depending on the sen- largely available commercial cameras (Table 1). Their
sors they carry on. Ongoing research is looking at the possible applications have been the focus of most
best solutions to integrate data collected from sensors on research for years due to their potential and low-cost
UAVs, ground sensors and other data sources for better operational requirements [74, 75].
management of punctual operations in the field, with a These sensors are used to calculate vegetation indices
particular focus on smart agriculture and big data man- such as the Green/Red Vegetation Index (GRVI), Green-
agement [68, 69]. ness Index (GI) and Excessive Greenness (ExG) with
Although UAVs systems do not offer the same territo- acceptable or high levels of accuracy [76, 77]. Also, RGB
rial coverage as satellites, they offer a spatial and temporal sensors have been increasingly used for machine learn-
ing techniques in object recognition, phenology, pathol-
ogies, and similar purposes. The typical workflow of
processing RGB images from UAVs for remote sensing is:
1. pre-flight planning, 2. flight and image acquisition, 3.
post-processing and indexes or dataset extrapolation [71].
Phase 1 is critical and essential to collect data of useful
quality for the purpose. In the pre-flight planning phase,
the parameters to consider are the definition of the study
area, the flight altitude, site topography, weather fore-
cast and local regulations for unmanned flights. In phase
2, it is recommended to keep the data flow sufficient to
store data and to check if the acquisition platform can
acquire the amount of data required. It could be possi-
ble to encounter I/O errors due to the inadequacy of the
platform with consequent loss of information or abor-
Fig. 2 Site-specific weed management (SSWM) scheme realized by tion of the mission. In phase 3, for RGB sensors, there is
drones and its economical and agro-ecological implications
no need to perform radiometric calibration, which is the
Esposito et al. Chem. Biol. Technol. Agric. (2021) 8:18 Page 4 of 11

Table 1 RGB cameras and their main specifications

Camera model Sensor type Sensor format Sensor Size [mm] Weight [kg] Price (approx.) [€]
and resolution
[Mpx]

Canon EOS 5d Mark IV CMOS 30.4 Full Frame 36.0 × 24.0 ca. 1.0 ca. 1000
Nikon D610 CMOS 24.3 Full Frame 36.0 × 24.0 ca. 1.250 ca. 1000
Sony Alpha 7R II CMOS 42 Full Frame Mirrorless 35.0 × 24.0 ca. 0.6 ca. 1200
Sony Alpha a6300 CMOS 24 Small Frame Mirrorless 23.5 × 15.6 ca. 0.8 ca. 800
Panasonic Lumix DMC GX8 CMOS 20 Small Frame Mirrorless 17.3 × 13 ca. 0.5 ca. 1000
Panasonic Lumix DMC GX80 DLMOS 16 Small Frame Mirrorless 17.3 × 13 ca. 0.5 ca. 500
DJI Phantom 4 Pro * CMOS 20 Small Frame 13.2 × 8.8 ca. 1.5 (with UAV) ca. 1500 (with UAV)
DJI Mavic 2 ­Proa CMOS 20 Small Frame 13.2 × 8.8 ca. 1.5 (with UAV) ca. 1500 (with UAV)
a *
UAV with already supplied camera. Payload not interchangeable

case when using multispectral and hyperspectral sensors. Multispectral sensors

RGB data can be used per se or to create a georeferenced The multispectral sensors are used for a wider range of
orthomosaic. In this case, the individual images are recti- calculations of vegetation indices as they can rely on a
fied, georeferenced using GPS data and stitched together higher number of radiometric bands. A comparison of
to form a single image (orthomosaic) covering the entire the most common multispectral sensors, specific for
study area. Orthomosaics can be generated either with UAV systems, is shown in Table 2.
RGB values as they are or after calculating the desired With multispectral sensors, the range of vegetation indi-
vegetation indices [77]. If RGB images are to be used ces that can be monitored is considerably extended com-
in machine learning algorithms, the workflow is differ- pared to those that can be calculated with only three RGB
ent [78–81]. In this case, it is necessary to collect a large bands. Moreover, the workflow has minor variations. For
dataset of images for the training and testing of the algo- these sensors, in phase 1, the radiometric calibration and
rithm [82]. This dataset may already be available from atmospheric correction phases are strictly required. Many
third-party sources, such as PlantVillage [83] or PlantDoc multispectral sensors, such as the Micasense RedEdge
[84]. Alternatively, it can be created from scratch if the series or the Parrot Sequoia + , have downwelling irradi-
purpose of the research is not covered by existing data- ance sensors and a calibrated reflectance panel to address
sets [85]. In this case, the acquisition, selection and pro- some of the requirements for radiometric calibration [87].
cessing of the images are critical, because the final dataset Due to a lower resolution of the sensors compared to RGB
can affect both the training and the use of the neural net- ones, a lower flight altitude and an adequate horizontal
work, with risks of producing biased results [86]. and vertical overlap of recorded images must be taken into

Table 2 Multispectral sensors and their main specifications

Camera model Resolution [Mpx] Spectral bands Ground sample distance Weight [kg] Price (approx.) [€]
[cm/px]

Micasense RedEdge-M 1280 × 960 (1.2 Mpx per EO Red, Green, Blue, Near- 8 (per band) at 120 m AGL ca. 0.180 ca. 5000
band) Infrared, Red Edge
Micasense RedEdge-MX 1280 × 960 (1.2 Mpx per EO Blue, green, red, red edge, 8 (per band) at 120 m AGL ca. 0.231 ca. 5000
band) near infrared (NIR)
Micasense Altum 2064 × 1544 (3.2 Mpx per EO EO: Blue, green, red, red 5.2 cm per pixel (per EO ca. 0.405 ca. 6000
band) 160 × 120 thermal edge, near-infrared (NIR) band) at 120 m AGL—
infrared LWIR: thermal infrared 81 cm per pixel (thermal)
8–14 µm at 120 m
TertaCam MCAW 6 1.3 6 user selectable narrow – ca. 0.550 ca. 17000
bands (450–1000 µm)
TetraCam ADC Lite 3.2 Green, Red, Near-Infrared – ca. 0.2 ca. 3000
(NIR)
TetraCam ADC Micro 3.2 Green, Red, Near-Infrared – ca. 0.09 ca. 3000
(NIR)
Parrot Sequoia + 1.2 Blue, Green, Red, Red Edge, – ca. 0.7 ca. 5000
Near-Infrared (NIR)
Esposito et al. Chem. Biol. Technol. Agric. (2021) 8:18 Page 5 of 11

account to obtain an adequate ground resolution for the to assess data quality, to correct radiance, and to generate
surveyed objective and to avoid missing data [88]. In phase a high-quality reflectance data-cube [93]. In phase 1, the
2, having a higher number of radiometric bands to record, planning must also be carried out in time and not only in
the dataflow will be higher so is critical to avoid I/O errors, space because, in addition to the spectrometric resolution,
missing data or mission failures [89]. Due to multi-lenses hyperspectral sensors have a temporal resolution due to the
nature of the sensors in phase 3, the data collected suffer different acquisition method. In phase 2, it should be con-
from the parallax problem. As a consequence, images have sidered that both images’ size and data flow are bigger than
to be rectified, georeferenced and must be stacked to gen- multispectral/RGB images. Moreover, these sensors may
erate a single image with different radiometric levels, and acquire a large amount of data, but the payload limitations
calibrated with the downwelling irradiance sensors data of UAVs may not allow the transport of adequate file storage
acquired during the flight [90]. After this procedure, it is systems. Phase 3 for hyperspectral images is critical: qual-
possible to generate a multispectral orthomosaic and then ity assessment is one of the critical issues of hyperspectral
calculate the requested indexes [91]. Multispectral images data and some problems associated with the quality of the
are also used in machine learning applications [80, 85, 92] images have not been completely overcome. Among those,
taking into account the multi-camera nature of sensors the stability of the sensor itself (due to the nature of UAV
and the different bands recorded. Thanks to the availabil- platforms) and the vibrations involved can comprise a good
ity of a higher number of radiometric bands, the machine calibration of the sensor. Subsequently, on post-processed
learning algorithms can be extended to not-visible recog- data, it is possible to calculate narrowband indices such as
nition such as early stage plant disease, field quality assess- chlorophyll absorption ratio index (CARI), greenness index
ment, soil water content, and more [91]. (GI), greenness vegetation index (GVI), modified chloro-
phyll absorption ratio index (MCARI), modified normalized
Hyperspectral sensors difference vegetation index (MNDVI), simple ratio (SR),
The hyperspectral sensors can record hundreds to thou- transformed chlorophyll absorption ratio index (TCARI),
sands of narrow radiometric bands, usually in visible and triangular vegetation index (TVI), modified vegetation
infrared ranges. To deal with hyperspectral applications, the stress ratio (MVSR), modified soil-adjusted vegetation index
choice of number and radiometric range of bands is criti- (MSAVI) and photochemical reflectance index (PRI) [94].
cal. Each band or combination of bands, being very narrow,
can detect a specific field characteristic. Each hyperspectral Applications of UAVs to weed management
sensor can detect only a certain number of bands, so the UAVs are ideal to identify weed patches. The main advan-
aim of survey must be very clear to choose the right sensor. tages of UAVs compared to UTVs are the shorter moni-
Although hyperspectral sensors have decreased in price in toring/surveying time they require and optimal control
recent years, they are still an important starting investment in the presence of obstacles, which is critical when work-
since they are much more expensive than RGB and multi- ing between crop rows [95]. In a few minutes, UAVs can
spectral sensors. In addition, they are heavier and bigger cover many hectares flying over the field, thus providing
than other sensors, often making their use on UAV systems the photographic material for weed patches identification
difficult and/or excessively onerous in terms of payload. [61]. These images are processed via deep neural network
Some of most used hyperspectral sensors in UAVs applica- [78], convolutional neural network, and object-based
tion and their main characteristics are shown in Table 3. image analysis [96, 97]. Based on a systematic review of
In this case, the workflow for radiometric calibration is the literature concerning weed identification by UAVs,
more complex compared to other sensors. Some calibration it can be concluded that mainly three types of cameras
methods needed for these sensors are derived from manned are used for weed patches identification: RGB, multispec-
aircraft hyperspectral platforms, based on artificial targets tral and hyperspectral cameras (Table 4). These cameras

Table 3 Hyperspectral sensors and their main characteristics

Camera model Lens Spectral range [µm] Spectral bands Weight [kg] Price (approx.) [€]
[number and µm]

CUBERT Snapshot + PAN 450–995 125 (8 µm) ca. 0.5 ca. 50000
Cornirg microHSI 410 SHARK CCD/CMOS 400–1000 300 (2 µm) ca. 0.7 –
Rikola Ltd. hyperspectral camera CMOS 500–900 40 (10 µm) ca. 0.6 ca. 40000
Specim-AISA KESTREL16 Push-broom 600–1640 350 (3 – 8 µm) ca. 2.5 –
Headwall Photonics InGaAs 900–1700 62 (12.9 µm) ca. 1.1 –
Micro-hyperspec X-series NIR
Esposito et al. Chem. Biol. Technol. Agric. (2021) 8:18 Page 6 of 11

Table 4 Weed patches identification by different types of camera (multispectral, RGB, hyperspectral)
Crop Weed (common name) Weed (scientific name) Type of camera Main results References

Palmer amaranth Amaranthus palmeri Hyperspectral camera Discriminate glyphosate- [110]

resistant from glyphosate-
sensitive weeds
Spotted knapweed Centaurea maculosa Hyperspectral camera Detection invasive species [111]
Babysbreath Gypsophila paniculata affecting forests, range-
lands, and pastures
Bunchgrass Phalaris minor RGB camera Identify different weeds [112]
Egyptian crowfoot grass Dactyloctenium aegyptium
False amaranth Digera arvensis
Awnless barnyard grass Echinochloa colona
Ragwort Jacobaea vulgaris (Senecio Multispectral camera Discriminate weeds in [113]
jacobaea) pastures
Buffel Grass Cenchrus ciliaris RGB camera Discriminate two different [114]
Spinifex Triodia sp. weeds
Beta vulgaris Kochia Bassia scoparia Hyperspectral camera Discriminate glyphosate [115]
Zea mays Marestail Conyza canadensis and dicamba resistant
Hordeum vulgare Common lambsquarters Chenopodium album genotypes from sensitive
Lens esculenta genotypes
Pisum sativum
Phaseolus vulgaris
Carthamus tinctorius
Cicer arietinum
Triticum spp. RGB camera Comparison of cereal geno- [116]
Triticosecale types
Beta vulgaris Weeds Multispectral camera Discriminate crop vs weeds [98]
Beta vulgaris Weeds Multispectral camera Discriminate crop vs weeds [85]
Beta vulgaris Thistle Cirsium arvense Multispectral camera Discriminate crop vs weeds [117]
Beta vulgaris Thistle Cirsium arvense Multispectral camera Discriminate crop vs weeds [101]
Wild buckwheat Fallopia convolvulus
Ryegrass Lolium multiflorum
Beta vulgaris Thistle Cirsium arvense Multispectral camera Discriminate crop vs weeds [112]
Cicer arietinum Weeds Hyperspectral camera Discriminate crop vs weeds [118]
Glycine max Palmer amaranth Amaranthus palmeri RGB camera Assessment of crop injury [102]
Barnyardgrass Echinochloa crus-galli Multispectral camera from dicamba
Large crabgrass Digitaria sanguinalis
Heliathus annuus Pigweed Amaranthus blitoides RGB camera Discriminate crop vs weeds [64]
Mustard Sinapis arvensis Multispectral camera
Bindweed Convolvulus arvensis L
Lambsquarters Chenopodium album L
Hordeum vulgare Thistle Cirsium arvense RGB camera Discriminate crop vs weeds [119]
Coltsfoot Tussilago farfara
Hordeum vulgare Thistle Cirsium arvense RGB camera Discriminate crop vs weeds [99]
Hordeum vulgare Thistle Cirsium arvense RGB camera Discriminate crop vs weeds [100]
Lactuca sativa Common groundsel Senecio vulgaris Multispectral camera Discriminate crops vs weeds [120]
Shepherd’s purse Capsella bursa pastoris
Sow thistle Sonchus spp.
Sorghum spp. Amaranth Amaranthus macrocarpus Hyperspectral camera Discriminate crop vs weeds [121]
Pigweed Portulaca oleracea
Barnyard grass Echinochloa crus-galli
Mallow E. colona
Nut grass Malva spp.
Fat Hen Cyperus rotundus
Chenopodium album
Triticum durum Wild oat Avena sterilis Multispectral camera Discriminate crop vs weeds [122]
Canarygrass Phalaris canariensis
Ryegrass Lolium rigidum
Triticum durum Wild oat Avena fatua Hyperspectral camera Discriminate crop vs weeds [105]
Canarygrass Phalaris canariensis Multispectral camera
Ryegrass Lolium rigidum
Esposito et al. Chem. Biol. Technol. Agric. (2021) 8:18 Page 7 of 11

Table 4 (continued)
Crop Weed (common name) Weed (scientific name) Type of camera Main results References

Triticum sp. Thistle Cirsium arvense RGB camera Discriminate crop vs weeds [99]
Triticum spp. Weeds Hyperspectral camera Discriminate crop vs weeds [118]
Vitis vinifera Bermuda grass Cynodon dactylon RGB camera Discriminate crop vs weeds [123]
Zea mays Weeds Multispectral camera Discriminate crop vs weeds [124]
Zea mays Common lambsquarters Chenopodium album Multispectral camera Discriminate monocotyle- [104]
Thistle Cirsium arvense dons (crops) vs dicotyle-
dons (weeds)
Zea mays Common lambsquarters Chenopodium album Multispectral camera Discriminate crop vs weeds [101]
Thistle Cirsium arvense
Zea mays Mat amaranth Amaranthus blitoides Multispectral camera Discriminate crop vs weeds [125]
Johnsongrass Sorghum halepense
These types of cameras can be assembled with drones for reaching SSWM purposes

are very similar in terms of information obtained for the strategies for weed management based on interspe-
purpose of weeds identification. Indeed, the three cam- cific crop–weed interactions [127–129]. Recent studies
era types can recognize weed patches with good accuracy demonstrate that some weed communities are actually
depending on flying altitude, camera resolution and UAV not detrimental to crop yield and quality [127, 128].
used. UAVs have been mainly tested on important crops In winter wheat cultivation, a highly diversified weed
such as Triticum spp., Hordeum vulgare, Beta vulgaris, community caused lower yield losses than a less diver-
Zea mays [98–101]. These are among the most cultivated sified one [129]. In soybean, through a combination of
crops worldwide and are highly susceptible to weed com- field experiments in which weed species were manipu-
petition especially in early phenological stages. In these lated in composition and abundance, it has been shown
crops, it was possible to identify several dicotyledonous that increasing levels of weed competition resulted in
weeds including Amaranthus palmeri, Chenopodium an increase in seed protein content without impairing
album and Cirsium arvense [102–104], as well as differ- yield [130].
ent monocotyledonous such as Phalaris spp., Avena spp. Most likely, the integration of known and emerging
and Lolium spp. [105, 106]. These weed species are wide- technologies in this field will greatly improve the sustain-
spread globally and can be a serious threat to different ability of weed control, following the SSWM approach.
crops [107, 108]. Therefore, the combined use of UAVs By image analysis, different machine learning techniques
and image processing technologies may contribute to will be able to provide a reliable overview of the level and
effectively control different weed species interfering with type of infestation. Specific algorithms can be trained to
the crops with relevant environmental benefits [28, 109]. manage weeds removal by Autonomous Weeding Robot
(AWR), via herbicide spray or mechanical means [131].
Conclusion Also, the creation of a specific weed images dataset is
The use of UAVs and machine learning techniques crucial to achieve this goal. This approach must neces-
allow for the identification of weed patches in a cul- sarily rely on a dataset of photographs taken in dedicated
tivated field with accuracy and can improve weed experimental fields, labeled in extended COCO/POCO
management sustainability [97]. Weed patches iden- (Common Objects in COntext/ Plant Objects in COn-
tification by UAVs can facilitate integrated weed man- text) format [86] and integrated with images from Plant-
agement (IWM), reducing both the selection pressure Village dataset [83] or other existing ones.
vs herbicide-resistant weeds and herbicides diffusion New insights on weed population dynamics and their
in the environment [64]. Recent research has shown competition with crops are needed in order to extend
that new technologies are able to discern single weed this approach to real agricultural contexts, so as to specif-
species in open fields [63, 106, 126]. If integrated with ically recognize and eliminate only harmful weed species.
weed management planning, this information gathered The overall objective is to overcome the consequences
via remote imaging analysis can contribute to sustain- of biological vacuum around the crop, which has been
ably improve weed management. In addition, imag- proved to be highly impacting for both biotic and the
ing analysis can help in the study of weed dynamics abiotic components of the environment [132, 133], with
in the field, as well as their interaction with the crop, long-term consequences on human safety on earth.
which both represent a necessary step to define new
Esposito et al. Chem. Biol. Technol. Agric. (2021) 8:18 Page 8 of 11

Abbreviations 15. Adkins S, Shabbir A. Biology, ecology and management of the invasive
IWM: Integrated Weed Management; UAV: Unmanned Aerial Vehicles; UTV: parthenium weed (Parthenium hysterophorus L.). Pest Manag Sci.
Unmanned Terrestrial Vehicles; SSWM: Specific Weed Management; RGB: Red 2014;70:1023–9.
Green Blue; VIS: Visible; GRVI: Green/Red Vegetation Index; GI: Greenness Index; 16. Tanveer A, Khaliq A. Mahajan G. Interference and management of par-
ExG: Excessive Greenness; GPS: Global Positioning System; CARI: Chlorophyll thenium: The world ’ s most important invasive weed. J Crop Prot; 2015.
Absorption Ratio Index; MCARI: Modified Chlorophyll Absorption Ratio p. 68.
Index; MNDVI: Modified Normalized Difference Vegetation Index; SR: Simple 17. Khan MR, Somvanshi VS, Rao U. Emerging nematode pest of rice, wheat
Ratio; TCARI: Transformed Chlorophyll Absorption Ratio Index; TVI: Triangular and onion. Rice Root-Knot Nematode Popular Kheti. 2017;3:53–5.
Vegetation Index; MVSR: Modified Vegetation Stress Ratio; MSAVI: Modified 18. Ersin A, Gargin S, Esitken A, Guzel NP, Atay AN, Altindal M, et al. The
Soil-Adjusted Vegetation Index; PRI: Photochemical Reflectance Index; AWR​ effect of weed competition on apple fruit quality. Not Bot Horti
: Autonomous Weeding Robot; COCO/POCO: Common Objects in COntext/ Agrobot Cluj-Napoca. 2017;45:120–5.
Plant Objects in COntext. 19. Yadav T, Chopra NK, Chopra NK, Kumar R, Soni PG. Assessment of
critical period of crop-weed competition in forage cowpea (Vigna
Acknowledgements unguiculata) and its effect on seed yield and quality. Indian J Agron.
Not applicable. 2018;63:124–7.
20. Hall CW. Vacuum treatment of milk. in: vacuum technology transac-
Authors’ contributions tions. Pergamon; 2013.
ME and MC wrote the original draft and produced all the tables and the 21. Sabra FS, Mahmoud MS. Utilization of herbicidal treatments to
figures. AM, FS and VC revised the text for the final version. VC and ME con- overcome weed problems in utilization of herbicidal treatments to
ceived the idea of writing the review. All authors read and approved the final overcome weed problems in peppermint (Mentha piperita L.) cultiva-
manuscript. tion under Egyptian conditions. Asian J Agric Food Sci. 2015;03:1–7.
22. Mantle P. Comparative ergot alkaloid elaboration by selected plecten-
Funding chymatic mycelia of Claviceps purpurea through sequential cycles of
This research was funded by MEDES Foundation. axenic culture and plant parasitism. Biology. 2020;9:41.
23. Kumar R, Katiyar R, Kumar S, Kumar T, Singh V. Lantana camara: an alien
Availability of data and materials weed, its impact on animal health and strategies to control. J Exp Biol
Not applicable. Agric Sci. 2016;4:321–37.
24. Denisow-Pietrzyk M, Pietrzyk Ł, Denisow B. Asteraceae species as
Competing interests potential environmental factors of allergy. Environ Sci Pollut Res.
The authors declare that they have no competing interests. 2019;26:6290–300.
25. Allan S, Shi B, Adkins SW. Impact of parthenium weed on human and
Received: 7 January 2021 Accepted: 15 February 2021 animal health. Parthenium Weed Biol Ecol Manag. 2018;7:105.
26. Korres NE, Burgos NR, Travlos I, Vurro M, Gitsopoulos TK, Varanasi
VK, et al. New directions for integrated weed management: mod-
ern technologies, tools and knowledge discovery. Adv Agron.
2019;155:243–319.
References 27. Partel V, Charan Kakarla S, Ampatzidis Y. Development and evaluation
1. Oerke EC. Crop losses to pests. Sci: J Agric; 2006. of a low-cost and smart technology for precision weed management
2. Guglielmini AC, Verdú AMC, Satorre EH. Competitive ability of five utilizing artificial intelligence. Comput Electron Agric. 2019;157:339–50.
common weed species in competition with soybean. Int J Pest Manag. 28. Hunter JE, Gannon TW, Richardson RJ, Yelverton FH, Leon RG.
2017;63:30–6. Integration of remote-weed mapping and an autonomous spraying
3. Kaur H, Brar SG, Shete PP. A REVIEW on different weed management unmanned aerial vehicle for site-specific weed management. Pest
approaches. Int J Curr Microbiol Appl Sci. 2019; 8: 2854. Manag Sci. 2020;76:1386–92.
4. Korav S, Ram V, Ray LIP, Krishnappa R, Singh NJ, Premaradhya N. Weed 29. Soltani N, Dille JA, Burke IC, Everman WJ, VanGessel MJ, Davis VM, et al.
pressure on growth and yield of groundnut (Arachis hypogaea L.) in Potential Corn yield losses from weeds in North America. Weed Technol.
Meghalaya, India. Int J Curr Microbiol Appl Sci. 2018;7:2852–8. 2016;30:979–84.
5. Thompson CR, Dille JA, Peterson DE. Weed Competition and Manage- 30. Llewellyn R, Ronning D, Ouzman J, Walker S, Mayfield A, Clarke M.
ment in Sorghum. In Sorghum. Ciampitti IA, Vara Prasad P, eds. 2019. Impact of Weeds on Australian Grain Production: the cost of weeds to
6. Zohaib A, Abbas T, Tabassum T. Weeds cause losses in field crops Australian grain growers and the adoption of weed management and
through allelopathy. Not Sci Biol. 2016;8:47–56. tillage practices Report for GRDC. CSIRO. 2016.
7. Cirillo V, Masin R, Maggio A, Zanin G. Crop-weed interactions in saline 31. Appleby AP, Müller F, Carpy S. Weed control. In Ullmann’s Encyclope-
environments. Eur J Agron. 2018;99:51–61. dia of Industrial Chemistry, (Ed.). https​://doi.org/10.1002/14356​007.
8. Hauvermale AL, Sanad MNME. Phenological plasticity of wild and a28_165
cultivated plants. Plant communities and their environment. London: 32. Swanton CJ, Nkoa R, Blackshaw RE. Experimental methods for crop-
IntechOpen; 2019. weed competition studies. Weed Sci. 2015;63:2–11.
9. Guo L, Qiu J, Li L-F, Lu B, Olsen K, Fan L. Genomic clues for crop-weed 33. EUSROSTAT. Statistics Eurostat. https​://ec.europ​a.eu/euros​tat/datab​
interactions and evolution. Trends Plant Sci. 2018;23:1102–15. rowse​r/view/aei_fm_salpe​st09/defau​lt/table​?lang=en.
10. Kumar A, Choudhary T, Das S, Meena SK. Weed Seed Bank: impacts and 34. Alexandratos N, Bruinsma J. World agriculture towards 2030/2050: the
management for future crop production BT—agronomic crops, Vol 2. 2012 revision. 2012.
Management Practices. Hasanuzzaman M, editor. 2019. p. 207–23. 35. Ray DK, Mueller ND, West PC, Foley JA. Yield trends are insufficient to
11. Smith JD, Dubois T, Mallogo R, Njau E-F, Tua S, Srinivasan R. Host range double global crop production by 2050. PLoS ONE. 2013;8:e66428.
of the invasive tomato pest tuta absoluta meyrick (Lepidoptera: 36. Mall RK, Gupta A, Sonkar G. 2—effect of climate change on agricultural
Gelechiidae) on solanaceous crops and weeds in Tanzania. Fla Entomol. crops. In: Dubey SK, Pandey A, Sangwan RSBT-CD in B and B, editors.
2019;101:573. Elsevier. 2017. p. 23–46.
12. Roshan P, Kulshreshtha A, Hallan V. Global weed-infecting geminiviruses 37. Westwood JH, Charudattan R, Duke SO, Fennimore SA, Marrone P,
BT-geminiviruses: impact, challenges and approaches. Berlin: Springer; Slaughter DC, et al. Weed management in 2050: perspectives on the
2019. p. 103–21. future of weed science. Weed Sci. 2018;66:275–85.
13. Srinivasan R, Cervantes FA, Alvarez JM. Aphid-borne virus dynamics in 38. Dong W, Liu E, Yan C, Tian J, Zhang H, Zhang Y. Impact of no tillage
the potato-weed pathosystem. Insect Pests Potato. 2013. vs. conventional tillage on the soil bacterial community structure in
14. Eshed N, Wahl I. Role of wild grasses in epidemics of powdery mildew. a winter wheat cropping succession in northern China. Eur J Soil Biol.
Phytopathology. 1975;65:57–63. 2017;80:35–42.
Esposito et al. Chem. Biol. Technol. Agric. (2021) 8:18 Page 9 of 11

39. Duke S, Heap I. Evolution of weed resistance to herbicides: what have 64. López-Granados F, Torres-Sánchez J, Serrano-Pérez A, de Castro AI,
we learned after 70 years. Biol Physiol Mol Biol Weeds. 2017;63–86. Mesas-Carrascosa Fco-J, Peña J-M. Early season weed mapping in sun-
40. Hicks HL, Comont D, Coutts SR, Crook L, Hull R, Norris K, et al. The fac- flower using UAV technology: variability of herbicide treatment maps
tors driving evolved herbicide resistance at a national scale. Nat Ecol against weed thresholds. Precis Agric. 2016;17:183–99.
Evol. 2018;2:529–36. 65. Jiménez-Brenes FM, López-Granados F, Torres-Sánchez J, Peña JM,
41. Büchi L, Wendling M, Amossé C, Necpalova M, Charles R. Importance Ramírez P, Castillejo-González IL, et al. Automatic UAV-based detection
of cover crops in alleviating negative effects of reduced soil tillage and of Cynodon dactylon for site-specific vineyard management. PLoS ONE.
promoting soil fertility in a winter wheat cropping system. Agric Ecosyst 2019.
Environ. 2018;256:92–104. 66. Daponte P, Vito LD, Glielmo L, Iannelli L, Liuzza D, Picariello F, et al. A
42. Jackson L, Calderon F, Steenwerth K, Scow K, Rolston D. Responses of review on the use of drones for precision agriculture. IOP Conf Ser Earth
soil microbial processes and community structure to tillage events and Environ Sci. 2019;275:012022.
implications for soil quality. Geoderma. 2003;114:305–17. 67. Radoglou-Grammatikis P, Sarigiannidis P, Lagkas T, Moscholios I. A com-
43. Muola A, Fuchs B, Laihonen M, Rainio K, Heikkonen L, Ruuskanen S, pilation of UAV applications for precision agriculture. Comput Netw.
et al. Risk in the circular food economy: glyphosate-based herbicide 2020;172:107148.
residues in manure fertilizers decrease crop yield. Sci Total Environ. 68. Mancini A, Frontoni E, Zingaretti P. Satellite and UAV data for Preci-
2020;750:141422. sion Agriculture Applications. In: 2019 International Conference on
44. Riemens MM, Van Der Weide RY, Bleeker PO, Lotz LAP. Effect of stale Unmanned Aircraft Systems (ICUAS). IEEE. 2019.
seedbed preparations and subsequent weed control in lettuce (cv. 69. Brook A, De Micco V, Battipaglia G, Erbaggio A, Ludeno G, Catapano I,
Iceboll) on weed densities. Weed Res. 2007;47:149–56. et al. A smart multiple spatial and temporal resolution system to sup-
45. Bajwa AA, Walsh M, Chauhan BS. Weed management using crop com- port precision agriculture from satellite images: proof of concept on
petition in Australia. Crop Prot. 2017;95:8–13. Aglianico vineyard. Remote Sens Environ. 2020;240:111679.
46. Swanton CJ, Weise SF. Integrated weed management: the rationale and 70. Watts AC, Ambrosia VG, Hinkley EA. Unmanned aircraft systems in
approach. Weed Technol. 1991;5:657–63. remote sensing and scientific research: classification and considerations
47. Young SL, Pitla SK, Van Evert FK, Schueller JK, Pierce FJ. Moving inte- of use. Remote Sens. 2012;4:1671–92.
grated weed management from low level to a truly integrated and 71. Manfreda S, McCabe ME, Miller PE, Lucas R, Madrigal VP, Mallinis G, et al.
highly specific weed management system using advanced technolo- On the use of unmanned aerial systems for environmental monitoring.
gies. Weed Res. 2017;57:1–5. Remote Sens. 2018;10:641.
48. Bonny S. Genetically modified herbicide-tolerant crops, weeds, and 72. Anderson K, Gaston KJ. Lightweight unmanned aerial vehicles will
herbicides: overview and impact. Environ Manage. 2016;57:31–48. revolutionize spatial ecology. Front Ecol Environ. 2013;11:138–46.
49. Rose MT, Cavagnaro TR, Scanlan CA, Rose TJ, Vancov T, Kimber S, et al. 73. Whitehead K, Hugenholtz CH. Remote sensing of the environment with
Impact of Herbicides on Soil Biology and Function. In: Sparks DL, editor. small unmanned aircraft systems (UASs), part 1: a review of progress
Adv Agro. Academic Press. 2016; pp. 133–220. and challenges. J Unmanned Veh Syst. 2014;02:69–85.
50. Shannon D, Clay DE, Sudduth KA. An introduction to precision agricul- 74. Torres-Sánchez J, Peña JM, de Castro AI, López-Granados F. Multi-tem-
ture. Precis Agric Basics. 2018. poral mapping of the vegetation fraction in early-season wheat fields
51. Mahlein AK. Plant disease detection by imaging sensors—parallels using images from UAV. Comput Electron Agric. 2014;103:104–13.
and specific demands for precision agriculture and plant phenotyping. 75. Jannoura R, Brinkmann K, Uteau D, Bruns C, Joergensen RG. Monitor-
Plant Dis. 2016;100:241–51. ing of crop biomass using true colour aerial photographs taken from a
52. Nawar S, Corstanje R, Halcro G, Mulla D, Mouazen AM. Delineation of remote controlled hexacopter. Biosyst Eng. 2015;129:341–51.
soil management zones for variable-rate fertilization: a review. Adv 76. Motohka T, Nasahara KN, Oguma H, Tsuchida S. Applicability of green-
Agron. 2017;143:175–245. red vegetation index for remote sensing of vegetation phenology.
53. Lama GFC, Errico A, Francalanci S, Solari L, Preti F, Chirico GB. Evalua- Remote Sens. 2010;2:2369–87.
tion of flow resistance models based on field experiments in a partly 77. Xue J, Su B. Significant remote sensing vegetation indices: a review of
vegetated reclamation channel. Geosciences. 2020;10:47. developments and applications. J Sens. 2017.
54. Gautam PV, Kushwaha HL, Kumar A, Kumar D. Mechatronics appli- 78. Crimaldi M, Cristiano V, De Vivo A, Isernia M, Ivanov P, Sarghini F. Neural
cation in precision sowing: a review. Int J Curr Microbiol Appl Sci. network algorithms for real time plant diseases detection using UAVs.
2019;8:1793–807. In: Coppola A, Di Renzo GC, Altieri G, D’Antonio P, editors. Innov Biosyst
55. Chlingaryan A, Sukkarieh S, Whelan B. Machine learning approaches Eng Sustain Agric For Food Prod. 2020. p. 827–35.
for crop yield prediction and nitrogen status estimation in precision 79. Pflanz M, Schirrmann M, Nordmeyer H. Drone based weed monitoring
agriculture: a review. Comput Electronics Agric. 2018;151:61–9. with an image feature classifier. Julius-Kühn-Archiv. 2018;458:379–84.
56. Murray SC. Optical sensors advancing precision in agricultural produc- 80. Etienne A, Saraswat D. Machine learning approaches to automate
tion. Photonics Spectra. 2017. weed detection by UAV based sensors. In: Autonomous air and ground
57. Dyshekov AI, Smirnov IG, Mirzaev MA, Shereuzhev MA. Principles of sensing systems for agricultural optimization and phenotyping IV.
functioning of the autonomous device for weed control for precision International Society for Optics and Photonics, 2019. p. 110080R.
agriculture. IOP Conf Ser Mater Sci Eng. 2020. 81. Wang S, Han Y, Chen J, Pan Y, Cao Y, Meng H, et al. A deep-learning-
58. Duhan JS, Kumar R, Kumar N, Kaur P, Nehra K, Duhan S. Nanotechnol- based low-altitude remote sensing algorithm for weed classification
ogy: the new perspective in precision agriculture. Biotechnol Rep. in ecological irrigation area. In: International Conference on Intelligent
2017;15:11–23. Technologies and Applications. Springer, Singapore, 2018. p. 451–460.
59. Mogili UR, Deepak BBVL. Review on application of drone systems in 82. Bah MD, Hafiane A, Canals R. Deep learning with unsupervised data
precision agriculture. Procedia Comput Sci. 2018;133:502–9. labeling for weed detection in line crops in UAV images. Remote Sens.
60. Raj R, Kar S, Nandan R, Jagarlapudi A. precision agriculture and 2018;10:1690.
unmanned aerial vehicles (UAVs) BT unmanned aerial vehicle: applica- 83. Hughes DP, Salathe M. An open access repository of images on plant
tions in agriculture and environment. In: Avtar R, Watanabe T, editors. health to enable the development of mobile disease diagnostics.
Cham: Springer International Publishing; 2020. ArXiv151108060 Cs. 2016. http://arxiv​.org/abs/1511.08060​
61. Krishna KR. Agricultural drones: a peaceful pursuit. Taylor Francis. 2018. 84. Singh D, Jain N, Jain P, Kayal P, Kumawat S, Batra N. PlantDoc: a dataset
62. John KN, Valentin V, Abdullah B, Bayat M, Kargar MH, Zargar M. Weed for visual plant disease detection. In: Proceedings of the 7th ACM IKDD
mapping technologies in discerning and managing weed infestation CoDS and 25th COMAD. 2020. p. 249–253.
levels of farming systems. Res Crops. 2020. 85. Sa I, Popović M, Khanna R, Chen Z, Lottes P, Liebisch F, et al. WeedMap: a
63. Lottes P, Khanna R, Pfeifer J, Siegwart R, Stachniss C. UAV-based crop large-scale semantic weed mapping framework using aerial multispec-
and weed classification for smart farming. Proc IEEE Int Conf Robot tral imaging and deep neural network for precision farming. Remote
Autom. 2017. Sens. 2018;10:1–25.
Esposito et al. Chem. Biol. Technol. Agric. (2021) 8:18 Page 10 of 11

86. Wspanialy P, Brooks J, Moussa M. An image labeling tool and agricul- 109. De Castro AI, Torres-Sánchez J, Peña JM, Jiménez-Brenes FM, Csillik
tural dataset for deep learning. ArXiv200403351 Cs Eess. 2020. http:// O, López-Granados F. An automatic random forest-OBIA algorithm
arxiv​.org/abs/2004.03351​ for early weed mapping between and within crop rows using UAV
87. Micasense Inc. Automatic Calibration Panel Detection (QR Mode). imagery. Remote Sens. 2018;10:285.
MicaSense Knowl. Base. 2020. https​://suppo​rt.micas​ense.com/hc/ 110. Reddy KN, Huang Y, Lee MA, Nandula VK, Fletcher RS, Thomson SJ, et al.
en-us/artic​les/36001​86187​74 Glyphosate-resistant and glyphosate-susceptible Palmer amaranth
88. Micasense Inc. Best practices: Collecting Data with MicaSense Sensors. (Amaranthus palmeri S. Wats.): hyperspectral reflectance properties of
MicaSense Knowl. Base. 2020. https​://suppo​rt.micas​ense.com/hc/ plants and potential for classification. Pest Manag Sci. 2014;70:1910–7.
en-us/artic​les/22489​3167. 111. Lass LW, Prather TS, Glenn NF, Weber KT, Mundt JT, Pettingill J. A
89. Laliberte AS, Goforth MA, Steele CM, Rango A. Multispectral review of remote sensing of invasive weeds and example of the early
remote sensing from unmanned aircraft: image processing work- detection of spotted knapweed (Centaurea maculosa) and babys-
flows and applications for rangeland environments. Remote Sens. breath (Gypsophila paniculata) with a hyperspectral sensor. Weed Sci.
2011;3:2529–51. 2005;53:242–51.
90. Barrero O, Perdomo SA. RGB and multispectral UAV image fusion for 112. Tiwari O, Goyal V, Kumar P, Vij S. An experimental set up for utilizing
Gramineae weed detection in rice fields. Agric: Precis; 2018. convolutional neural network in automated weed detection. Proc 2019
91. Candiago S, Remondino F, De Giglio M, Dubbini M, Gattelli M. Evaluat- 4th Int Conf Internet Things Smart Innov Usages IoT-SIU 2019. 2019;1–6.
ing multispectral images and vegetation indices for precision farming 113. Tay JYL, Erfmeier A, Kalwij JM. Reaching new heights: can drones
applications from UAV images. Remote Sens. 2015;7:4026. replace current methods to study plant population dynamics? Plant
92. Mazzia V, Comba L, Khaliq A, Chiaberge M, Gay P. UAV and machine Ecol. 2018;219:1139–50.
learning based refinement of a satellite-driven vegetation index for 114. Sandino J, Gonzalez F. A Novel Approach for invasive weeds and veg-
precision agriculture. Sensors. 2020;20:2530. etation surveys using UAS and artificial intelligence. 2018 23rd Int Conf
93. Brook A, Ben-Dor E. Supervised Vicarious Calibration (SVC) of Methods Models Autom Robo. 2018.
multi-source hyperspectral remote-sensing data. Remote Sens. 115. Scherrer B, Sheppard J, Jha P, Shaw JA. Hyperspectral imaging and
2015;7:6196–223. neural networks to classify herbicide-resistant weeds. J Appl Remote
94. Adao T, Hruska J, Padua L, Bessa J, Peres E, Morais R, et al. Hyperspectral Sens. 2019;13:1.
imaging: a review on UAV-based sensors, data processing and applica- 116. Svensgaard J, Jensen SM, Westergaard JC, Nielsen J, Christensen S,
tions for agriculture and forestry. Remote Sens. 2017;9:1110. Rasmussen J. Can reproducible comparisons of cereal genotypes be
95. Hassanein M, El-Sheimy N. An efficient weed detection procedure using generated in field experiments based on UAV imagery using RGB
low-cost UAV imagery system for precision agriculture applications. Int cameras? Eur J Agron. 2019;106:49–57.
Arch Photogramm Remote Sens Spat Inf Sci ISPRS Arch. 2018;42:181–7. 117. Garcia-Ruiz FJ, Wulfsohn D, Rasmussen J. Sugar beet (Beta vulgaris L.)
96. Tsouros DC, Bibi S, Sarigiannidis PG. A review on UAV-based applica- and thistle (Cirsium arvensis L.) discrimination based on field spectral
tions for precision agriculture. Inf Switz. 2019;10:349. data. Biosyst Eng. 2015;139:1–15.
97. Maes WH, Steppe K. Perspectives for remote sensing with unmanned 118. Shapira U, Herrmann I, Karnieli A, Bonfil JD. Weeds detection by ground-
aerial vehicles in precision agriculture. Trends Plant Sci. 2019;24:152–64. level hyperspectral data. ISPRS Archive Vol. XXXVIII. 2010;38:27–33.
98. Fawakherji M, Potena C, Bloisi DD, Imperoli M, Pretto A, Nardi D. UAV 119. Rasmussen J, Nielsen J, Garcia-Ruiz F, Christensen S, Streibig JC. Poten-
Image based crop and weed distribution estimation on embedded tial uses of small unmanned aircraft systems (UAS) in weed research.
GPU boards BT computer analysis of images and patterns. Berlin: Weed Res. 2013;53:242–8.
Springer; 2019. 120. Slaughter DC, Giles DK, Fennimore SA, Smith RF. Multispectral machine
99. Rasmussen J, Nielsen J, Streibig JC, Jensen JE, Pedersen KS, Olsen SI. vision identification of lettuce and weed seedlings for automated weed
Pre-harvest weed mapping of Cirsium arvense in wheat and barley with control. Weed Technol. 2008;22:378–84.
off-the-shelf UAVs. Precis Agric. 2019;20:983–99. 121. Norasma N, Ya C, George D. Spectral discrimination of weeds using
100. Franco C, Guada C, Rodríguez JT, Nielsen J, Rasmussen J, Gómez D, et al. hyperspectral. 2013.
Automatic detection of thistle-weeds in cereal crops from aerial RGB 122. López-Granados F, Peña-Barragán JM, Jurado-Expósito M, Francisco-
images BT information processing and management of uncertainty in Fernández M, Cao R, Alonso-Betanzos A, et al. Multispectral classifi-
knowledge-based systems. Applications. 2018. p. 441–52. cation of grass weeds and wheat (Triticum durum) using linear and
101. Louargant M, Jones G, Faroux R, Paoli JN, Maillot T, Gée C, et al. nonparametric functional discriminant analysis and neural networks.
Unsupervised classification algorithm for early weed detection in Weed Res. 2008;48:28–37.
row-crops by combining spatial and spectral information. Remote Sens. 123. de Castro AI, Peña JM, Torres-Sánchez J, Jiménez-Brenes F, López-Gra-
2018;10:1–18. nados F. Mapping Cynodon dactylon in vineyards using UAV images for
102. Huang Y, Reddy KN, Fletcher RS, Pennington D. UAV low-altitude remote site-specific weed control. Adv Anim Biosci. 2017;8:267–71.
sensing for precision weed management. Weed Technol. 2018;32:2–6. 124. Fawakherji M, Potena C, Bloisi DD, Imperoli M, Pretto A, Nardi D. UAV
103. Hansen KD, Garcia-Ruiz F, Kazmi W, Bisgaard M, La Cour-Harbo A, Ras- image based crop and weed distribution estimation on embedded
mussen J, et al. An autonomous robotic system for mapping weeds in GPU boards. Commun Comput Inf Sci. 2019;1089:100–8.
fields. IFAC Proc Vol IFAC-Pap. 2013;8:217–24. 125. Peña JM, Torres-Sánchez J, de Castro AI, Kelly M, López-Granados F.
104. Louargant M, Villette S, Jones G, Vigneau N, Paoli JN, Gée C. Weed Weed mapping in early-season maize fields using object-based analysis
detection by UAV: simulation of the impact of spectral mixing in multi- of unmanned aerial vehicle (UAV) images. PLoS ONE. 2013;8:1–11.
spectral images. Precis Agric. 2017;1–20. 126. Pantazi X-E, Moshou D, Bravo C. Active learning system for weed
105. López-Granados F, Jurado-Expósito M, Peña-Barragán JM, García-Torres species recognition based on hyperspectral sensing. Biosyst Eng.
L. Using remote sensing for identification of late-season grass weed 2016;146:193–202.
patches in wheat. Weed Sci. 2006;54:346–53. 127. Gibson DJ, Young BG, Wood AJ. Can weeds enhance profitability?
106. Sanders JT, Everman WJ, Austin R, Roberson GT, Richardson RJ. Weed Integrating ecological concepts to address crop-weed competition and
species differentiation using spectral reflectance land image classifica- yield quality. J Ecol. 2017;105:900–4.
tion. ProcSPIE. 2019. 128. Millar KDL, Ebbs SD, Gibson DJ, Wood AJ, Young BG. Evaluation of
107. Davis S, Mangold J, Menalled F, Orloff N, Miller Z, Lehnhoff E. A Meta- physiological parameters for the prediction of seed yield and quality for
analysis of Canada Thistle (Cirsium arvense) management. Weed Sci. soybean (Glycine max) plants grown in the presence of weed competi-
2018;66:548–57. tion. Plant Biosyst. 2011;145:1–11.
108. Alshallash KS. Germination of weed species (Avena fatua, Bromus 129. Adeux G, Vieren E, Carlesi S, Bàrberi P, Munier-Jolain N, Cordeau S.
catharticus, Chenopodium album and Phalaris minor) with implications Mitigating crop yield losses through weed diversity. Nat Sustain.
for their dispersal and control. Ann Agric Sci. 2018;63:91–7. 2019;2:1018–26.
Esposito et al. Chem. Biol. Technol. Agric. (2021) 8:18 Page 11 of 11

130. Millar K, Gibson DJ, Young BG, Wood AJ. Impact of interspecific compe- 133. Erisman JW, van Eekeren N, de Wit J, Koopmans C, Cuijpers W, Oerle-
tition on seed development and quality of five soybean cultivars. Aust J mans N, et al. Agriculture and biodiversity: a better balance benefits
Exp Agric. 2007;47:1455–9. both. AIMS Agric Food. 2016;1:157–74.
131. Talaviya T, Shah D, Patel N, Yagnik H, Shah M. Implementation of artificial
intelligence in agriculture for optimisation of irrigation and application
of pesticides and herbicides. Artif Intell Agric. 2020;4:58–73. Publisher’s Note
132. Lanz B, Dietz S, Swanson T. The expansion of modern agriculture and Springer Nature remains neutral with regard to jurisdictional claims in pub-
global biodiversity decline: an integrated assessment. Ecol Econ. lished maps and institutional affiliations.
2018;144:260–77.