Ci a ion: Blanco-Ca mona, P.;
Baeza-Mo eno, L.; Hidalgo-Fo , E.;
Ma ín-Clemen e, R.; González-
Ca ajal, R.; Muñoz-Cha e o, F. AIoT
in Ag icul u e: Sa egua ding C ops
om Pes and Disease Th ea s.
Senso s 2023,23, 9733. h ps://
doi.o g/10.3390/s23249733
Academic Edi o : Ra aele B uno
Recei ed: 7 No embe 2023
Re ised: 5 Decembe 2023
Accep ed: 6 Decembe 2023
Published: 10 Decembe 2023
Copy igh : © 2023 by he au ho s.
Licensee MDPI, Basel, Swi ze land.
This a icle is an open access a icle
dis ibu ed unde he e ms and
condi ions o he C ea i e Commons
A ibu ion (CC BY) license (h ps://
c ea i ecommons.o g/licenses/by/
4.0/).
senso s
A icle
AIoT in Ag icul u e: Sa egua ding C ops om Pes and
Disease Th ea s
Ped o Blanco-Ca mona 1, Lucía Baeza-Mo eno 1, Edua do Hidalgo-Fo 1,* , Rubén Ma ín-Clemen e 2,
Ramón González-Ca ajal 1and Fe nando Muñoz-Cha e o 1
1Depa men o Elec onic Enginee ing, Uni e si y o Se ille, 41092 Se ille, Spain; [email p o ec ed] (P.B.-C.);
[email p o ec ed] (L.B.-M.); [email p o ec ed] (R.G.-C.); [email p o ec ed] (F.M.-C.)
2Depa men o Signal P ocessing and Communica ions, Uni e si y o Se ille, 41092 Se ille, Spain;
[email p o ec ed]
*Co espondence: [email p o ec ed]
Abs ac :
A signi ican p opo ion o he wo ld’s ag icul u al p oduc ion is los o pes s and diseases.
To mi iga e his p oblem, an AIoT sys em o he ea ly de ec ion o pes and disease isks in c ops
is p oposed. I p esen s a sys em based on low-powe and low-cos senso nodes ha collec
en i onmen al da a and ansmi i once a day o a se e ia a NB-IoT ne wo k. In addi ion, he
senso nodes use indi idual, e ainable and upda able machine lea ning algo i hms o assess he
isk le el in he c op e e y 30 min. I a isk is de ec ed, en i onmen al da a and he isk le el a e
immedia ely sen . Addi ionally, he sys em enables wo ypes o no i ica ion: email and lashing LED,
p o iding online and o line isk no i ica ions. As a esul , he sys em was deployed in a eal-wo ld
en i onmen and he powe consump ion o he senso nodes was cha ac e ized, alida ing hei
longe i y and he co ec unc ioning o he isk de ec ion algo i hms. This allows he a me o know
he s a us o hei c op and o ake ea ly ac ion o add ess hese h ea s.
Keywo ds:
In e ne o Things (IoT); Wi eless Senso Ne wo k (WSN); NB-IoT; sma ag icul u e;
A i icial In elligence (AI)
1. In oduc ion
Each yea , up o 40% o he wo ld’s ag icul u al p oduc i i y is los o he a ages o
pes s and diseases [
1
]. A signi ican p opo ion o his s agge ing loss can be a ibu ed o
he lack o imely de ec ion and iden i ica ion o hese h ea s. In he pas , his p oblem
was isually de ec ed, by obse ing changes in he appea ance o he c op. Howe e ,
his is a edious and ime-consuming p ocess when ying o assess he condi ion o each
indi idual plan .
To da e in human his o y, he e ha e been ou ag icul u al e olu ions [
2
]. The
ea lies o hese was known as adi ional ag icul u e, whe e indi iduals solely elied
on manual labo and animal powe o end o hei c ops. In he mid-20 h cen u y, he
second was cha ac e ized by he in oduc ion o powe ed machine y sys ems, allowing a
dec ease in manual labo and an inc ease in p oduc i i y, coupled wi h he use o e ilize s
and pes icides. The la e 20 h cen u y ma ked he in oduc ion o au oma ed ag icul u e
and moni o ing c ops. Finally, he mos ecen e olu ion inco po a es he undamen al
p inciples o Indus y 4.0, in ol ing ad ances such as A i icial In elligence, he In e ne
o Things and Big Da a, in oducing new applica ions: sma me e ing [
3
,
4
] and a i icial
in elligence o op imize c op p oduc i i y [
5
,
6
], e ec i ely add essing issues ela ed o pes s
and diseases wi hou eso ing o con en ional pes icide use. O he no able de elopmen s
include in elligen i iga ion me hods [
7
], moni o ing sys ems o manage sma e a ms [
8
],
he moni o ing o wa e consump ion le els [
9
] and e o s o educe bo h en i onmen al
impac and esou ce deple ion [10,11].
Senso s 2023,23, 9733. h ps://doi.o g/10.3390/s23249733 h ps://www.mdpi.com/jou nal/senso s
Senso s 2023,23, 9733 2 o 19
Wi h ega d o pes s and disease isk de ec ion sys ems, se e al solu ions ha e been
p oposed. These include compu e ision echniques ha analyze images o signs o
pes in es a ion [
12
] o disease mani es a ion [
13
]. The e has also been esea ch in o plan
diseases using ae ial images aken by unmanned ae ial ehicles (UAVs) [
14
]. In addi ion,
some sys ems ha e been de eloped o collec en i onmen al da a o aid bo h pes and
disease de ec ion [
15
]. Some app oaches e en combine di e en echnologies o add ess
his p oblem [16].
The In e ne o Things (IoT) and A i icial In elligence (AI) ha e p o en o be in alu-
able ools in his espec . By applying hese wo echnologies oge he , a ema kable usion
known as he A i icial In elligence o Things (AIoT) has eme ged, showing immense
po en ial. Some o he mos in e es ing ea u es o his new pa adigm include:
•
The a ailabili y o new low-cos senso s, which make i cheape o c ea e wi eless
de ices and allow o a mul i ude o senso s adap ed o he desi ed needs.
•
Wi eless communica ions, such as Na owBand-IoT (NB-IoT) [
17
], which allow high
co e age in u al en i onmen s.
•Low-powe mic op ocesso sys ems, which ensu e he longe i y o AIoT de ices.
•The abili y o analyze si ua ions and make decisions.
The main objec i e o his pape ocuses on he success ul implemen a ion o a low-
powe and low-cos AIoT pes and disease isk de ec ion sys em. I should be no ed
ha se e al exis ing sys ems ely on image analysis, al hough wi h inaccu a e esul s
when applied o la ge a eas, while o he s use expensi e and high-main enance UAVs o
achie e he desi ed esul s. In addi ion, some sys ems collec en i onmen al da a using AI
algo i hms o acili a e he de elopmen o ene gy-e icien and cos -e ec i e solu ions.
This pape s a s wi h an o e iew o he en i e sys em, explaining how i wo ks. This
is ollowed by an explana ion o he mos impo an componen s o he sys em, namely
senso nodes and he se e . In addi ion, emphasis is placed on he de elopmen o he
algo i hm used o he de ec ion o pes and disease isks, as well as on he no i ica ion
sys em in he case o isk de ec ion. Then, i p o ides a desc ip ion o he expe imen al
esul s ob ained and compa isons wi h o he simila sys ems, o inally summa ize he
conclusions ob ained om his wo k.
2. Ma e ials and Me hods
2.1. Sys em O e iew
2.1.1. Sys em A chi ec u e
The sys em a chi ec u e is shown in Figu e 1whe e o ange a ows ep esen uplink
and pu ple ones downlink communica ions.
The sys em is composed o h ee blocks:
•
Senso nodes: The senso nodes a e asked wi h collec ing da a om en i onmen al
senso s. Using he NB-IoT ne wo k, his in o ma ion is hen wi elessly ansmi ed
o he se e . In addi ion, hese senso s a e e alua ed in eal ime using a machine
lea ning algo i hm. This is used o de e mine i any pes o disease h ea s a e eme g-
ing in he ag icul u al landscape. In cases whe e isks a e de ec ed, no i ica ions
a e immedia ely sen o he a me , ale ing hem o po en ial p oblems in speci ic
a eas o hei c op. A decision ee algo i hm was chosen o his AIoT sys em be-
cause o i s compa ibili y wi h low-powe nodes and i s inna e simplici y o human
comp ehension.
•
NB-IoT: The NB-IoT module ac s as a wi eless communica ion ool wi hin he NB-IoT
ne wo k, which is managed by mobile ope a o s. This module es ablishes di ec
connec ions and is designed o p o ide senso nodes wi h a long ange and longe
ba e y li e due o i s IoT-compa ible in as uc u e. This compa ibili y makes i an
essen ial componen o una ended de ices deployed in ag icul u al en i onmen s,
b inging signi ican bene i s o hese sys ems.
Senso s 2023,23, 9733 3 o 19
•
Se e : The se e ecei es da a om he senso nodes and s o es i in a da abase. This
da a is hen used o con inuously ain he machine lea ning algo i hm. I is impo an
o no e ha he se e also manages impo an isk ale s gene a ed by he senso
nodes. These ale s a e immedia ely sen o he a m owne , allowing hem o eac in
a imely way and ake he necessa y measu es o p e en any po en ial loss o c op o
p oduc ion yield.
Senso s 2023, 23, x FOR PEER REVIEW 3 o 20
Figu e 1. Sys em a chi ec u e.
The sys em is composed o h ee blocks:
• Senso nodes: The senso nodes a e asked wi h collec ing da a om en i onmen al
senso s. Using he NB-IoT ne wo k, his in o ma ion is hen wi elessly ansmi ed o
he se e . In addi ion, hese senso s a e e alua ed in eal ime using a machine lea n-
ing algo i hm. This is used o de e mine i any pes o disease h ea s a e eme ging
in he ag icul u al landscape. In cases whe e isks a e de ec ed, no i ica ions a e im-
media ely sen o he a me , ale ing hem o po en ial p oblems in speci ic a eas o
hei c op. A decision ee algo i hm was chosen o his AIoT sys em because o i s
compa ibili y wi h low-powe nodes and i s inna e simplici y o human comp ehen-
sion.
• NB-IoT: The NB-IoT module ac s as a wi eless communica ion ool wi hin he NB-
IoT ne wo k, which is managed by mobile ope a o s. This module es ablishes di ec
connec ions and is designed o p o ide senso nodes wi h a long ange and longe
ba e y li e due o i s IoT-compa ible in as uc u e. This compa ibili y makes i an
essen ial componen o una ended de ices deployed in ag icul u al en i onmen s,
b inging signi ican bene i s o hese sys ems.
• Se e : The se e ecei es da a om he senso nodes and s o es i in a da abase.
This da a is hen used o con inuously ain he machine lea ning algo i hm. I is im-
po an o no e ha he se e also manages impo an isk ale s gene a ed by he
senso nodes. These ale s a e immedia ely sen o he a m owne , allowing hem o
eac in a imely way and ake he necessa y measu es o p e en any po en ial loss
o c op o p oduc ion yield.
2.1.2. Desc ip ion o Sys em Func ionali y
The senso ne wo k has h ee di e en use unc ionali ies:
• Da a logge (Figu e 2): Senso nodes collec empe a u e, humidi y, and ain all da a
e e y 30 min. This in o ma ion is s o ed in hei memo y. E e y day, hese da a a e
sen o he se e . Sending da a e e y day was chosen as a comp omise be ween he
equi emen o immediacy whils minimizing powe consump ion, hus inc easing
ba e y li e. A daily ansmission equency is adequa e o moni o c ops. I o some
eason he senso node de ec s a p oblema ic si ua ion, i would immedia ely send
Figu e 1. Sys em a chi ec u e.
2.1.2. Desc ip ion o Sys em Func ionali y
The senso ne wo k has h ee di e en use unc ionali ies:
•
Da a logge (Figu e 2): Senso nodes collec empe a u e, humidi y, and ain all da a
e e y 30 min. This in o ma ion is s o ed in hei memo y. E e y day, hese da a a e
sen o he se e . Sending da a e e y day was chosen as a comp omise be ween he
equi emen o immediacy whils minimizing powe consump ion, hus inc easing
ba e y li e. A daily ansmission equency is adequa e o moni o c ops. I o some
eason he senso node de ec s a p oblema ic si ua ion, i would immedia ely send all
he s o ed da a in o ma ion wi h an app op ia e wa ning, as can be seen in he ‘ isk
assessmen and no i ica ion’ unc ionali y.
Senso s 2023, 23, x FOR PEER REVIEW 4 o 20
all he s o ed da a in o ma ion wi h an app op ia e wa ning, as can be seen in he
‘ isk assessmen and no i ica ion’ unc ionali y.
Figu e 2. Da a logge unc ionali y.
• Risk e alua ion and no i ica ion (Figu e 3): Each ime en i onmen al da a a e col-
lec ed in he da a logge unc ionali y, he isk le el o he a ea is e alua ed. By de-
aul , he e is no isk and he sys em ac s as a da a logge . Howe e , i i de ec s a
isk, i immedia ely sends a isk ale along wi h all he in o ma ion om he senso s.
In his case, he se e ecei es he ale and manages he necessa y no i ica ions so
ha he a me is awa e ha he e is a p oblem on he a m. In addi ion, he de ice
ins alled on he a m has isual wa ning ligh s ha lash a di e en equencies
when hey de ec a ce ain le el o isk in he a ea. This unc ionali y enables a low-
powe and once-a-day solu ion ha is also capable o de ec ing apid changes and
imp o ing sys em esponsi eness.
Figu e 3. Risk e alua ion and no i ica ion.
• T aining and upda ing o he machine lea ning algo i hm (Figu e 4): Once a mon h,
he se e uses he collec ed en i onmen al da a o e ain each machine lea ning al-
go i hm. These algo i hms, which a e speci ically ained o each senso node, a e
upda ed hanks o he Fi mwa e O e The Ai (FOTA) ea u e. This allows he algo-
i hm o adap o di e en e ain condi ions, depending on whe e he de ice is in-
s alled.
Figu e 2. Da a logge unc ionali y.
•
Risk e alua ion and no i ica ion (Figu e 3): Each ime en i onmen al da a a e collec ed
in he da a logge unc ionali y, he isk le el o he a ea is e alua ed. By de aul ,
he e is no isk and he sys em ac s as a da a logge . Howe e , i i de ec s a isk, i
Senso s 2023,23, 9733 4 o 19
immedia ely sends a isk ale along wi h all he in o ma ion om he senso s. In his
case, he se e ecei es he ale and manages he necessa y no i ica ions so ha he
a me is awa e ha he e is a p oblem on he a m. In addi ion, he de ice ins alled
on he a m has isual wa ning ligh s ha lash a di e en equencies when hey
de ec a ce ain le el o isk in he a ea. This unc ionali y enables a low-powe and
once-a-day solu ion ha is also capable o de ec ing apid changes and imp o ing
sys em esponsi eness.
Senso s 2023, 23, x FOR PEER REVIEW 4 o 20
all he s o ed da a in o ma ion wi h an app op ia e wa ning, as can be seen in he
‘ isk assessmen and no i ica ion’ unc ionali y.
Figu e 2. Da a logge unc ionali y.
• Risk e alua ion and no i ica ion (Figu e 3): Each ime en i onmen al da a a e col-
lec ed in he da a logge unc ionali y, he isk le el o he a ea is e alua ed. By de-
aul , he e is no isk and he sys em ac s as a da a logge . Howe e , i i de ec s a
isk, i immedia ely sends a isk ale along wi h all he in o ma ion om he senso s.
In his case, he se e ecei es he ale and manages he necessa y no i ica ions so
ha he a me is awa e ha he e is a p oblem on he a m. In addi ion, he de ice
ins alled on he a m has isual wa ning ligh s ha lash a di e en equencies
when hey de ec a ce ain le el o isk in he a ea. This unc ionali y enables a low-
powe and once-a-day solu ion ha is also capable o de ec ing apid changes and
imp o ing sys em esponsi eness.
Figu e 3. Risk e alua ion and no i ica ion.
• T aining and upda ing o he machine lea ning algo i hm (Figu e 4): Once a mon h,
he se e uses he collec ed en i onmen al da a o e ain each machine lea ning al-
go i hm. These algo i hms, which a e speci ically ained o each senso node, a e
upda ed hanks o he Fi mwa e O e The Ai (FOTA) ea u e. This allows he algo-
i hm o adap o di e en e ain condi ions, depending on whe e he de ice is in-
s alled.
Figu e 3. Risk e alua ion and no i ica ion.
•
T aining and upda ing o he machine lea ning algo i hm (Figu e 4): Once a mon h,
he se e uses he collec ed en i onmen al da a o e ain each machine lea ning
algo i hm. These algo i hms, which a e speci ically ained o each senso node,
a e upda ed hanks o he Fi mwa e O e The Ai (FOTA) ea u e. This allows he
algo i hm o adap o di e en e ain condi ions, depending on whe e he de ice
is ins alled.
Senso s 2023, 23, x FOR PEER REVIEW 5 o 20
Figu e 4. T aining and upda ing machine lea ning algo i hm unc ionali y.
2.2. Senso s Nodes
The senso nodes we e speci ically de eloped o his wo k and hey a e made up o
wo di e en boa ds. On he one hand, he Co e Boa d is esponsible o he p ocessing
and con ol o he pe iphe als, and on he o he hand, he Senso Boa d ca ies ou he
senso iza ion. The co e o he i s boa d is a STM32L152RE mic op ocesso (ST mic oe-
lec onics manu ac u e ), a mic op ocesso based on an ul a-low powe ARM Co ex M3,
wi h a clock o 32 MHz and able o ope a e be ween 1.65 and 3.6 V. This mic ocon olle
allows us o un a eal- ime ope a ing sys em (F eeRTOS), which helps o concu en ly
pe o m senso da a collec ion, machine lea ning algo i hm execu ion and da a ansmis-
sion.
As addi ional pe iphe als o he mic op ocesso , he co e boa d (Figu e 5) is
equipped wi h:
• EEPROM memo y o s o age o da a acqui ed by he senso s.
• A SIM7080 NB-IoT anscei e (SIMCOM manu ac u e ), used o send he collec ed
da a o he se e . I communica es wi h he mic op ocesso based on AT commands
and a s andby powe consump ion o 3 µA.
• SIM ca d, equi ed o connec ion o he mobile NB-IoT ope a o ne wo k.
• P og amming in e ace (JTAG).
• Access o he I2C, SPI and UART communica ion in e aces. These in e aces a e es-
sen ial o managing wo k lows.
Figu e 5. Co e Boa d.
ECONATUR ag icul u al expe ise was used o iden i y which senso s a e he mos
app op ia e o iden i y pes and disease isks, ob aining he ollowing senso s:
Figu e 4. T aining and upda ing machine lea ning algo i hm unc ionali y.
2.2. Senso s Nodes
The senso nodes we e speci ically de eloped o his wo k and hey a e made up o
wo di e en boa ds. On he one hand, he Co e Boa d is esponsible o he p ocessing and
con ol o he pe iphe als, and on he o he hand, he Senso Boa d ca ies ou he senso iza-
ion. The co e o he i s boa d is a STM32L152RE mic op ocesso (ST mic oelec onics
manu ac u e ), a mic op ocesso based on an ul a-low powe ARM Co ex M3, wi h a
clock o 32 MHz and able o ope a e be ween 1.65 and 3.6 V. This mic ocon olle allows
us o un a eal- ime ope a ing sys em (F eeRTOS), which helps o concu en ly pe o m
senso da a collec ion, machine lea ning algo i hm execu ion and da a ansmission.
As addi ional pe iphe als o he mic op ocesso , he co e boa d (Figu e 5) is
equipped wi h:
•EEPROM memo y o s o age o da a acqui ed by he senso s.
•
A SIM7080 NB-IoT anscei e (SIMCOM manu ac u e ), used o send he collec ed
da a o he se e . I communica es wi h he mic op ocesso based on AT commands
and a s andby powe consump ion o 3 µA.
•SIM ca d, equi ed o connec ion o he mobile NB-IoT ope a o ne wo k.
Senso s 2023,23, 9733 5 o 19
•P og amming in e ace (JTAG).
•
Access o he I2C, SPI and UART communica ion in e aces. These in e aces a e
essen ial o managing wo k lows.
Senso s 2023, 23, x FOR PEER REVIEW 5 o 20
Figu e 4. T aining and upda ing machine lea ning algo i hm unc ionali y.
2.2. Senso s Nodes
The senso nodes we e speci ically de eloped o his wo k and hey a e made up o
wo di e en boa ds. On he one hand, he Co e Boa d is esponsible o he p ocessing
and con ol o he pe iphe als, and on he o he hand, he Senso Boa d ca ies ou he
senso iza ion. The co e o he i s boa d is a STM32L152RE mic op ocesso (ST mic oe-
lec onics manu ac u e ), a mic op ocesso based on an ul a-low powe ARM Co ex M3,
wi h a clock o 32 MHz and able o ope a e be ween 1.65 and 3.6 V. This mic ocon olle
allows us o un a eal- ime ope a ing sys em (F eeRTOS), which helps o concu en ly
pe o m senso da a collec ion, machine lea ning algo i hm execu ion and da a ansmis-
sion.
As addi ional pe iphe als o he mic op ocesso , he co e boa d (Figu e 5) is
equipped wi h:
• EEPROM memo y o s o age o da a acqui ed by he senso s.
• A SIM7080 NB-IoT anscei e (SIMCOM manu ac u e ), used o send he collec ed
da a o he se e . I communica es wi h he mic op ocesso based on AT commands
and a s andby powe consump ion o 3 µA.
• SIM ca d, equi ed o connec ion o he mobile NB-IoT ope a o ne wo k.
• P og amming in e ace (JTAG).
• Access o he I2C, SPI and UART communica ion in e aces. These in e aces a e es-
sen ial o managing wo k lows.
Figu e 5. Co e Boa d.
ECONATUR ag icul u al expe ise was used o iden i y which senso s a e he mos
app op ia e o iden i y pes and disease isks, ob aining he ollowing senso s:
Figu e 5. Co e Boa d.
ECONATUR ag icul u al expe ise was used o iden i y which senso s a e he mos
app op ia e o iden i y pes and disease isks, ob aining he ollowing senso s: empe a u e,
humidi y and c op p ecipi a ion. Ma ke esea ch o di e en senso s was ca ied ou .
These we e chosen based on he ollowing c i e ia: accu acy and ene gy consump ion.
On he ma ke , empe a u e and humidi y senso s a e usually included in a single
senso . Table 1shows he di e en senso s e alua ed acco ding o he cha ac e is ics
men ioned abo e.
Table 1. E alua ion o empe a u e and humidi y senso s.
Senso Resolu ion (bi s) Temp. Tole ance (◦C) Hum. Tole ance (%) Powe
Consump ion (µA) P ice (€)
SHTC3 16 ±0.2 ◦C±2% Ac i e: 430
Sleep: 0.6 1.05
HDC1080DMBR 14 ±0.2 ◦C±2% Ac i e: 710
Sleep: 0.1 3.84
SHT30-DIS 16 ±0.2 ◦C±2% Ac i e: 600
Sleep: 0.2 2.97
SHT20 14 ±0.2 ◦C±3% Ac i e: 300
Sleep: 0.15 4.57
As all senso s had e y simila ole ances and he esolu ion o 14 o 16 bi s was
no signi ican o ou applica ion, i was decided o p io i ize he powe consump ion,
choosing he SHT20 senso , as i has he lowes powe consump ion.
Rega ding he ain gauge, hey a e usually associa ed wi h comple e wea he s a ions,
bu a low-cos , low-consump ion solu ion was needed. The e o e, he WH-SP-RG ain
gauge wi h pulsed ou pu was chosen, which minimized he cos and gua an eed low
ene gy consump ion.
Thus, he senso node was equipped wi h a empe a u e and humidi y senso SHT20
and a WH-SP-RG ain gauge.
As can be seen in Figu e 6, he senso boa d designed has some non-welded compo-
nen s ha a e in ended o addi ional asks o be ca ied ou du ing he o e all p ojec .
Senso s 2023,23, 9733 6 o 19
Senso s 2023, 23, x FOR PEER REVIEW 6 o 20
empe a u e, humidi y and c op p ecipi a ion. Ma ke esea ch o di e en senso s was
ca ied ou . These we e chosen based on he ollowing c i e ia: accu acy and ene gy con-
sump ion.
On he ma ke , empe a u e and humidi y senso s a e usually included in a single
senso . Table 1 shows he di e en senso s e alua ed acco ding o he cha ac e is ics men-
ioned abo e.
Table 1. E alua ion o empe a u e and humidi y senso s.
Senso
Resolu ion (bi s)
Temp. Tole ance
(°C)
Hum. Tole ance (%)
Powe Consump ion (µA)
P ice
(€)
SHTC3
16
±0.2 °C
±2%
Ac i e: 430
Sleep: 0.6
1.05
HDC1080DMBR
14
±0.2 °C
±2%
Ac i e: 710
Sleep: 0.1
3.84
SHT30-DIS
16
±0.2 °C
±2%
Ac i e: 600
Sleep: 0.2
2.97
SHT20
14
±0.2 °C
±3%
Ac i e: 300
Sleep: 0.15
4.57
As all senso s had e y simila ole ances and he esolu ion o 14 o 16 bi s was no
signi ican o ou applica ion, i was decided o p io i ize he powe consump ion, choos-
ing he SHT20 senso , as i has he lowes powe consump ion.
Rega ding he ain gauge, hey a e usually associa ed wi h comple e wea he s a-
ions, bu a low-cos , low-consump ion solu ion was needed. The e o e, he WH-SP-RG
ain gauge wi h pulsed ou pu was chosen, which minimized he cos and gua an eed low
ene gy consump ion.
Thus, he senso node was equipped wi h a empe a u e and humidi y senso SHT20
and a WH-SP-RG ain gauge.
As can be seen in Figu e 6, he senso boa d designed has some non-welded compo-
nen s ha a e in ended o addi ional asks o be ca ied ou du ing he o e all p ojec .
Figu e 6. Senso Boa d.
The senso node, comp ised o he wo assembled boa ds, he ba e y (LSP33600-20F),
he an enna and he senso s, is shown in Figu e 7:
Figu e 6. Senso Boa d.
The senso node, comp ised o he wo assembled boa ds, he ba e y (LSP33600-20F),
he an enna and he senso s, is shown in Figu e 7:
Senso s 2023, 23, x FOR PEER REVIEW 7 o 20
Figu e 7. Senso Node.
A he i mwa e le el, d i e s we e c ea ed o manage each senso and manage he
ex e nal pe iphe als, as well as all he F eeRTOS asks equi ed.
The abo e ask diag am (Figu e 8), whe e blue a ows ep esen queues and o ange
ones, RTOS semapho es, can be desc ibed as ollows: Ini ialisa ion is esponsible o ini-
ializing all he en i onmen al senso s and he NB-IoT anscei e , as well as de ec ing
whe he FOTA is equi ed. The Measu eTempHum and Measu eRain all asks hen collec
he empe a u e and humidi y, and ain all, espec i ely, and pass hem o he Ma-
chineLea ningAlgo i hms ask, which is esponsible o de ec ing whe he he e a e pes
and/o disease isks using he decision ee algo i hms. I he e is no isk, he sys em goes
in o s andby mode, and i he e is a isk, i sends i o e NB-IoT ia he ModemManage .
In addi ion, da a is pe iodically sen , whe he he e is a isk o no . Addi ionally, he e is
a debugging ask ( Debug), wa chdog managemen ( Res eshIWDG), low powe manage-
men ( Po Sup essTicksAndSleep) and managemen o di e en ala ms (RTCAla mA-
BCallback). Thanks o his mul i- asking sys em, a high le el o sys em eliabili y is
achie ed, enabling long- e m deploymen by allowing ac ions o be aken in he e en o
sys em ailu e, such as: Res eshIWDG ask es a s he sys em when he mic op ocesso
goes in o an unknown s a e, Measu eTempHum and Measu eRain all asks un sequences
o eboo and/o shu down and powe up senso s in he e en o a lack o communica ion
wi h hem, and ModemManage ask s o es he da a o be sen in he u u e in case he
ne wo k is una ailable a ha momen .
Figu e 7. Senso Node.
A he i mwa e le el, d i e s we e c ea ed o manage each senso and manage he
ex e nal pe iphe als, as well as all he F eeRTOS asks equi ed.
The abo e ask diag am (Figu e 8), whe e blue a ows ep esen queues and o ange
ones, RTOS semapho es, can be desc ibed as ollows: Ini ialisa ion is esponsible o ini-
ializing all he en i onmen al senso s and he NB-IoT anscei e , as well as de ec ing
whe he FOTA is equi ed. The Measu eTempHum and Measu eRain all asks hen collec
he empe a u e and humidi y, and ain all, espec i ely, and pass hem o he Machine-
Senso s 2023,23, 9733 7 o 19
Lea ningAlgo i hms ask, which is esponsible o de ec ing whe he he e a e pes and/o
disease isks using he decision ee algo i hms. I he e is no isk, he sys em goes in o
s andby mode, and i he e is a isk, i sends i o e NB-IoT ia he ModemManage . In
addi ion, da a is pe iodically sen , whe he he e is a isk o no . Addi ionally, he e is a de-
bugging ask ( Debug), wa chdog managemen ( Res eshIWDG), low powe managemen
( Po Sup essTicksAndSleep) and managemen o di e en ala ms (RTCAla mA-BCallback).
Thanks o his mul i- asking sys em, a high le el o sys em eliabili y is achie ed, enabling
long- e m deploymen by allowing ac ions o be aken in he e en o sys em ailu e,
such as: Res eshIWDG ask es a s he sys em when he mic op ocesso goes in o an
unknown s a e, Measu eTempHum and Measu eRain all asks un sequences o eboo
and/o shu down and powe up senso s in he e en o a lack o communica ion wi h
hem, and ModemManage ask s o es he da a o be sen in he u u e in case he ne wo k
is una ailable a ha momen .
Senso s 2023, 23, x FOR PEER REVIEW 8 o 20
Figu e 8. F eeRTOS asks diag am.
2.3. Decision T ee
A decision ee algo i hm has he pa icula i y ha i is human eadable and compu-
a ionally simple, so i can be easily implemen ed in he senso node. I is essen ial o be
able o ha e an easy- o-in eg a e algo i hm because he nodes will use edge compu ing o
de ec i he e is any ype o isk and quickly no i y i .
The algo i hm was ini ially ained using da abases om me eo ological s a ions,
such as ha o IFAPA [18], and using he knowledge o ag icul u al expe s o labeling
pes and disease isks. The inpu ea u es o he algo i hm a e cu en empe a u e, max-
imum day ime empe a u e, cu en humidi y, maximum day ime humidi y and cu en
ain all. Finally, se e al decision ees we e adjus ed o de ec a ce ain numbe o pes s
and diseases a di e en dep hs and sizes. Figu e 9 shows a shallow decision ee o un-
de s and how i wo ks h ough i o he condi ions wi h di e en en i onmen al a iables
a e in ol ed. In gene al, he p edic ion o he algo i hm is a ce ain pes /in es a ion isk
le el classi ied in o h ee ypes: low, medium and high. An example is shown in Figu e 9
o a decision ee e alua ion wi h Tamb = 24, HR = 10, HR_Max = 40, assuming a medium
le el o isk o diseases.
Once he senso was ins alled, he da a collec ed we e used by he se e o u he
ain he algo i hm and imp o e i . In his way, i was possible o upda e he nodes wi h
hese emo ely ained algo i hms. The se e is p epa ed o e ain he algo i hms, which
can be emo ely upda ed on he senso nodes a any ime. The e- ainabili y o he algo-
i hms gi es he ad an age o making new algo i hms ha can be adap ed o he
Figu e 8. F eeRTOS asks diag am.
2.3. Decision T ee
A decision ee algo i hm has he pa icula i y ha i is human eadable and compu-
a ionally simple, so i can be easily implemen ed in he senso node. I is essen ial o be
able o ha e an easy- o-in eg a e algo i hm because he nodes will use edge compu ing o
de ec i he e is any ype o isk and quickly no i y i .
The algo i hm was ini ially ained using da abases om me eo ological s a ions, such
as ha o IFAPA [
18
], and using he knowledge o ag icul u al expe s o labeling pes
and disease isks. The inpu ea u es o he algo i hm a e cu en empe a u e, maximum
day ime empe a u e, cu en humidi y, maximum day ime humidi y and cu en ain all.
Finally, se e al decision ees we e adjus ed o de ec a ce ain numbe o pes s and diseases
Senso s 2023,23, 9733 8 o 19
a di e en dep hs and sizes. Figu e 9shows a shallow decision ee o unde s and how
i wo ks h ough i o he condi ions wi h di e en en i onmen al a iables a e in ol ed.
In gene al, he p edic ion o he algo i hm is a ce ain pes /in es a ion isk le el classi ied
in o h ee ypes: low, medium and high. An example is shown in Figu e 9o a decision
ee e alua ion wi h Tamb = 24, HR = 10, HR_Max = 40, assuming a medium le el o isk
o diseases.
Senso s 2023, 23, x FOR PEER REVIEW 9 o 20
eme gence o new pes s and/o diseases based on he labels es ablished by he ag icul u al
expe s.
Figu e 9. Decision ee o disease isk e alua ion. Tamb = ambien empe a u e, HR = ela i e hu-
midi y and HR_Max = ela i e maximum humidi y in he las 24 h. The algo i hm ou pu s an as-
sessmen o isk, i.e., BAJO (low), MEDIO (medium) and ALTO (high).
The h eshold o he decision ees will change o e ime. Thus, each senso node will
ha e i s own speci ic decision ee. Depending on he dis ibu ion o he senso nodes, hey
will be able o ob ain di e en empe a u e, humidi y and ain all alues, especially in la ge
c ops, so he pes and disease isk assessmen will change. This ea u e allows he algo i hm
o adap o he ype o e ain, ype o c op, ype o pes , ype o disease, clima e, e c.
The senso node, h ough he algo i hm, sends no i ica ions as soon as a change in isk
le el is de ec ed, as opposed o wai ing an en i e day o send he in o ma ion o he se e .
2.4. Se e
Wi h ega d o s o age, da a explo a ion, analysis and isualiza ion o he in o -
ma ion ecei ed by he emo e nodes, a mic ose ices a chi ec u e is a ailable o gua an-
ee scalabili y and isola ion be ween he di e en applica ions in ol ed in he p ocess.
Figu e 10 shows a diag am wi h he basic mic ose ices a chi ec u e in ol ed in he in-
o ma ion managemen sys em, as well as o he in e media e middlewa e esou ces use-
ul o he co ec de elopmen o he sys em (secu i y laye s, load balancing manage-
men , e e se p oxy, e c.).
Figu e 9.
Decision ee o disease isk e alua ion. Tamb = ambien empe a u e, HR = ela i e
humidi y and HR_Max = ela i e maximum humidi y in he las 24 h. The algo i hm ou pu s an
assessmen o isk, i.e., BAJO (low), MEDIO (medium) and ALTO (high).
Once he senso was ins alled, he da a collec ed we e used by he se e o u he
ain he algo i hm and imp o e i . In his way, i was possible o upda e he nodes
wi h hese emo ely ained algo i hms. The se e is p epa ed o e ain he algo i hms,
which can be emo ely upda ed on he senso nodes a any ime. The e- ainabili y
o he algo i hms gi es he ad an age o making new algo i hms ha can be adap ed
o he eme gence o new pes s and/o diseases based on he labels es ablished by he
ag icul u al expe s.
The h eshold o he decision ees will change o e ime. Thus, each senso node will
ha e i s own speci ic decision ee. Depending on he dis ibu ion o he senso nodes,
hey will be able o ob ain di e en empe a u e, humidi y and ain all alues, especially
in la ge c ops, so he pes and disease isk assessmen will change. This ea u e allows
he algo i hm o adap o he ype o e ain, ype o c op, ype o pes , ype o disease,
clima e, e c.
The senso node, h ough he algo i hm, sends no i ica ions as soon as a change in isk
le el is de ec ed, as opposed o wai ing an en i e day o send he in o ma ion o he se e .
2.4. Se e
Wi h ega d o s o age, da a explo a ion, analysis and isualiza ion o he in o ma ion
ecei ed by he emo e nodes, a mic ose ices a chi ec u e is a ailable o gua an ee scala-
bili y and isola ion be ween he di e en applica ions in ol ed in he p ocess. Figu e 10
shows a diag am wi h he basic mic ose ices a chi ec u e in ol ed in he in o ma ion
managemen sys em, as well as o he in e media e middlewa e esou ces use ul o he
co ec de elopmen o he sys em (secu i y laye s, load balancing managemen , e e se
p oxy, e c.).
Senso s 2023,23, 9733 9 o 19
Senso s 2023, 23, x FOR PEER REVIEW 10 o 20
Figu e 10. Se e unc ional block diag am.
The i s block belongs o he senso nodes deployed in he c op, which send senso
da a ia he NB-IoT ne wo k o he se e . The da a a e sen e e y day unless a ce ain
ype o isk has been de ec ed.
The se e module, which o ms he co e o his sec ion, is composed o mic ose ices
(applica ions) ha ope a e in isola ion bu a e able o communica e wi h each o he . The
main se ices deployed in he se e a e:
• Da a s o age: I o e sees he s o age o all da a collec ed by he senso nodes, bo h
o isualiza ion and o ain he decision ee algo i hm.
• Da a p ocessing: This mic ose ice pe o ms he whole algo i hm e aining p oce-
du e, which is execu ed e e y mon h and gene a es new algo i hms o be upda ed,
h ough FOTA, in senso nodes.
• Email se ice: I is esponsible o building and sending weekly e-mail no i ica ions
on he s a us o he en i onmen al a iables sensed, and o sending no i ica ions o
pes and/o disease isk de ec ion.
• F on end: The on end mic ose ice is esponsible o p o iding he necessa y web
in e ace so ha he use can easily access he con en s s o ed on he se e .
• Backend: I suppo s he on end ope a ions, making any da a i equi es a ailable
and ensu ing ha all eques s a e secu e.
• O he mic ose ices: This pape is ocused on he main ea u es o he sys em, how-
e e , o he mic ose ices a e execu ed in pa allel, such as secu i y laye s and load
balancing managemen .
Finally, he e is a block o he p esen a ion o in o ma ion o he use and suppo
o he expo o o iginal and sys em-p ocessed da a, as well as impo ing ex e nal da a.
The web in e ace is p esen ed in Figu e 11. I shows he geog aphical a ea whe e
some o he e minal nodes c ea ed ha e been loca ed. O he ea u es in eg a ed in he
websi e a e isualiza ion (Figu e 12) and he expo /impo o he da a.
Figu e 10. Se e unc ional block diag am.
The i s block belongs o he senso nodes deployed in he c op, which send senso
da a ia he NB-IoT ne wo k o he se e . The da a a e sen e e y day unless a ce ain ype
o isk has been de ec ed.
The se e module, which o ms he co e o his sec ion, is composed o mic ose ices
(applica ions) ha ope a e in isola ion bu a e able o communica e wi h each o he . The
main se ices deployed in he se e a e:
•
Da a s o age: I o e sees he s o age o all da a collec ed by he senso nodes, bo h o
isualiza ion and o ain he decision ee algo i hm.
•
Da a p ocessing: This mic ose ice pe o ms he whole algo i hm e aining p ocedu e,
which is execu ed e e y mon h and gene a es new algo i hms o be upda ed, h ough
FOTA, in senso nodes.
•
Email se ice: I is esponsible o building and sending weekly e-mail no i ica ions
on he s a us o he en i onmen al a iables sensed, and o sending no i ica ions o
pes and/o disease isk de ec ion.
•
F on end: The on end mic ose ice is esponsible o p o iding he necessa y web
in e ace so ha he use can easily access he con en s s o ed on he se e .
•
Backend: I suppo s he on end ope a ions, making any da a i equi es a ailable
and ensu ing ha all eques s a e secu e.
•
O he mic ose ices: This pape is ocused on he main ea u es o he sys em, howe e ,
o he mic ose ices a e execu ed in pa allel, such as secu i y laye s and load balancing
managemen .
Finally, he e is a block o he p esen a ion o in o ma ion o he use and suppo o
he expo o o iginal and sys em-p ocessed da a, as well as impo ing ex e nal da a.
The web in e ace is p esen ed in Figu e 11. I shows he geog aphical a ea whe e
some o he e minal nodes c ea ed ha e been loca ed. O he ea u es in eg a ed in he
websi e a e isualiza ion (Figu e 12) and he expo /impo o he da a.
On he web page, he da a acqui ed om he di e en senso s a e displayed. Fo
ins ance, in Figu e 12, he g aphs o he empe a u e da a in deg ees Celsius (uppe cha )
and humidi y in pe cen age (lowe cha ) o e a pe iod o one week a e shown.
The abili y o expo he da a was in oduced o allow hem o be displayed on o he
media and/o o c ea e new machine lea ning algo i hms ha can be used in he senso
nodes, as well as o he unc ionali ies.
Senso s 2023,23, 9733 16 o 19
Senso s 2023, 23, x FOR PEER REVIEW 17 o 20
Figu e 18. Almond ee wi h a ew aphids on some o i s lea es.
Figu e 19. Almond ee comple ely in es ed wi h aphids.
Figu e 19. Almond ee comple ely in es ed wi h aphids.
Du ing hese mon hs, he algo i hm was emo ely upda ed o check he co ec pe -
o mance o his unc ionali y, wi h sa is ac o y esul s. Fu he mo e, i was e i ied ha
all he de ices co ec ly wo ked, wi hou he need o eplace ba e ies, no p oblems we e
de ec ed in co e age ailu e, and no samples we e los .
4. Compa a i e S udy and Discussion
In his a icle, a compa a i e s udy o he de eloped sys em wi h p e iously p oposed
sys ems was ca ied ou . Table 3compa es he main cha ac e is ics o he AIoT sys ems.
Table 3. Compa a i e wi h o he s AIoT Sys ems.
Fea u es Ching-Ju
Chen e al. [21]
Oli ie
Debauche e al. [22]N. Ma e ne e al. [23] T. T. Win e al. [24] This Wo k
Senso s
Tempe a u e,
humidi y, mobile
phone, UAV
Tempe a u e,
humidi y, ba ome e ,
ain, gauge, .. .
Ai empe a u e, ai
humidi y, CO2,
illumina ion in ensi y,
soil empe a u e, soil
humidi y, soil
mois u e, lea
we ness
Tempe a u e,
humidi y, p essu e,
sunligh le el,
wa e le el
Tempe a u e,
humidi y,
ain gauge
AI algo i hm
execu ion Se e Node Se e Se e Node
De ec pes s Yes Yes Yes No Yes
De ec diseases No Yes Yes Yes (wi h images) Yes
Ale s ia in e ne
Yes Yes No No Yes
Visual ale s No No No No Yes
E alua ion pe iod Each 1 h Each 5 min Each 30 min - Each 30 min
T ansmission
in e al Each 1 h Each 5 min Each 30 min - Each day
Ne wo k LoRa LoRa ZigBee LoRa NB-IoT
Au onomy - - - - 17 yea s
Senso s 2023,23, 9733 17 o 19
Compa ing hese sys ems, he use o edge compu ing in [
22
] and his wo k s ands ou ,
as he sys em does no depend on he se e o he execu ion o he algo i hm. Addi ionally,
he sys em by Chen e al. lacks disease de ec ion. While N. Ma e ne e al. [
23
] used ZigBee,
he h ee o he sys ems use LoRa, and his wo k employs NB-IoT. NB-IoT, in con as o
ZigBee and LoRa, has he ad an age ha i is a sel -managed ne wo k, being ope a ed by
he mobile ope a o s. In his case, his ne wo k is a ailable in almos 100% o he Spanish
e i o y, wi h e y compe i i e p ices.
The algo i hms p esen ed in his wo k a e e ainable, so hey ha e high lexibili y
o changes such as new pes s, new diseases, he ype o c op, he ype o e ain, e c. I
he algo i hm de ec s an inc eased isk, he sys em ale s he a me . An impo an poin
is he p esence o isual ale s in he sys em p esen ed in his a icle, which is essen ial
o he a me no o be comple ely dependen on he In e ne . I is unde s ood ha
di e en a me s will wan di e en me hods o no i ica ion. As all ala ms a e s o ed on he
se e , he de eloped sys em allows he in eg a ion o any ype o ala m, ei he physical
(e.g., audible ala m), in o ma ion echnology (e.g., Teleg am bo ) o mobile (e.g., SMS).
This wo k pe mi s he use o edge compu ing echniques, minimizing he powe
consump ion o he senso node, and does educe he esponse ime when no i ying o
a ce ain isk in he c op. The e alua ion ime is 30 min, close o he o he sys ems. On
he o he hand, he ansmission in e al inc eases by up o one ansmission pe day. In
addi ion, he numbe o senso s has been minimized, only using hose ha expe s ha e
conside ed necessa y o his applica ion. This esul s in a compe i i ely p iced, low-powe
and low-cos de ice.
5. Conclusions
This wo k demons a es ha i is possible o de elop a pes and disease isk de ec ion
sys em using AIoT echnology by de eloping low-powe and low-cos senso nodes. To
his aim, all sys em unc ionali ies (da a logge , isk assessmen and no i ica ion, aining
and upda ing o he machine lea ning algo i hm) we e alida ed wi h sa is ac o y esul s
in a eal-wo ld en i onmen .
In addi ion, a decision ee algo i hm is p oposed capable o de ec ing he pes and
disease isk le els in c ops. ECONATUR ag icul u al expe s ha e de e mined ha i is
possible o assess pes and disease isks based on c op empe a u e, humidi y and ain all
measu emen s, minimizing he numbe o senso s equi ed and he e o e educing he cos
o he senso node.
Thanks o con inuous da a acquisi ion, his sys em can con inuously imp o e he
algo i hm and be emo ely upda ed. Ha ing a senso nodes ne wo k in he c op allows o
he de ec ion o he loca ion whe e he isk has eme ged and will make i possible o s udy
he e olu ion o pes s and diseases in he c op. The ad an age o e aining and upda ing
algo i hms allows o an adap a ion o new pes s and diseases, as well as di e en ypes o
e ain and/o c ops.
This sys em employs edge compu ing echniques, inc easing ansmission in e als
ia NB-IoT and uses a minimized senso sui , allowing o a low-powe , low-cos solu ion.
The sys em also p o ides a high le el o esponsi eness o changes ha may be de imen al
o he c op, as unde isk- ee condi ions he ansmission a e may be one day, bu i a isk
is de ec ed, he ansmission is immedia e. The isks de ec ed a e communica ed o he
a me by means o ale s. Two a e p oposed in his wo k: isual ale s and e-mail ale s.
Howe e , hanks o he ecep ion o isk epo s, he se e is eady o implemen any o he
ype o ale ha a a me may eques .
Wi h he esul s ob ained, he easibili y o AIoT sys ems used o de ec pes s and
diseases in c ops was demons a ed, p o iding ea ly isk wa nings o c op issues. In his
way, i allows he a me o quickly ac and minimize annual p oduc ion losses.
Au ho Con ibu ions:
Concep ualiza ion, P.B.-C., R.M.-C., R.G.-C. and F.M.-C.; me hodology,
P.B.-C., L.B.-M., E.H.-F., R.M.-C., R.G.-C. and F.M.-C.; so wa e, P.B.-C., L.B.-M. and E.H.-F.; al-
ida ion P.B.-C., L.B.-M., E.H.-F. and R.M.-C.; o mal analysis, P.B.-C. and L.B.-M.; in es iga ion, P.B.-C.,
Senso s 2023,23, 9733 18 o 19
L.B.-M., E.H.-F. and R.M.-C.; esou ces P.B.-C., L.B.-M., E.H.-F., R.M.-C., R.G.-C. and F.M.-C.; da a
cu a ion, P.B.-C., L.B.-M. and E.H.-F.; w i ing—o iginal d a p epa a ion P.B.-C., L.B.-M. and E.H.-F.;
w i ing— e iew and edi ing, P.B.-C., R.M.-C. and R.G.-C.; isualiza ion, P.B.-C., L.B.-M. and E.H.-F.;
supe ision, P.B.-C., R.G.-C. and F.M.-C. All au ho s ha e ead and ag eed o he published e sion
o he manusc ip .
Funding:
The au ho s would like o hank he Spanish minis ies o inno a ion and science o und-
ing his esea ch unde G an PID2019-107258RB-C31 unded by MCIN/AEI/10.13039/501100011033
and Indus y, T ade and Tou ism unde esea ch p ojec IAg i ( e . AEI-010500-2020-188).
Ins i u ional Re iew Boa d S a emen : No applicable.
In o med Consen S a emen : No applicable.
Da a A ailabili y S a emen : Da a a e con ained wi hin he a icle.
Acknowledgmen s: Au ho s would like o hank he ag icul u al company ECONATUR.
Con lic s o In e es : The au ho s decla e no con lic o in e es .
Re e ences
1.
FAO Launches 2020 as he UN’s In e na ional Yea o Plan Heal h. 2019. A ailable online: h ps://www. ao.o g/news/s o y/
en/i em/1253551/icode/ (accessed on 2 Sep embe 2023).
2.
Liu, Y.; Ma, X.; Hancke, G.P.; Abu-Mah ouz, A.M. F om Indus y 4.0 o ag icul u e 4.0: Cu en s a us, enabling echnologies, and
esea ch challenges. IEEE T ans. Ind. In o m. 2021,17, 4322–4334. [C ossRe ]
3.
Ullo, S.L.; Sinha, G.R. Ad ances in Sma En i onmen Moni o ing Sys ems Using IoT and Senso s. Senso s
2020
,20, 3113.
[C ossRe ] [PubMed]
4.
Jawad, H.M.; No din, R.; Gha ghan, S.K.; Jawad, A.M.; Ismail, M. Ene gy-E icien Wi eless Senso Ne wo ks o P ecision
Ag icul u e: A e iew. Senso s 2017,17, 1781. [C ossRe ] [PubMed]
5.
Thanga aj, R.; Anandamu ugan, S.; Pandiyan, P.; Kaliappan, V.K. A i icial in elligence in oma o lea Disease De ec ion: A
Comp ehensi e Re iew and discussion. J. Plan Dis. P o . 2021,129, 469–488. [C ossRe ]
6.
Ayed, R.B.; Hanana, M. A i icial in elligence o imp o e he ood and ag icul u e sec o . J. Food Qual.
2021
,2021, 5584754.
[C ossRe ]
7.
Ga cía, L.; Pa a, L.; Jiménez, J.M.; Llo e , J.; Lo enz, P. IoT-Based Sma I iga ion Sys ems: An o e iew on he ecen ends on
senso s and IoT sys ems o i iga ion in p ecision ag icul u e. Senso s 2020,20, 1042. [C ossRe ] [PubMed]
8.
Almalki, F.A.; Sou iene, B.O.; Alsamhi, S.H.; Sakli, H. A Low-Cos pla o m o en i onmen al sma a ming moni o ing sys em
based on IoT and UAVs. Sus ainabili y 2021,13, 5908. [C ossRe ]
9.
Soh ZH, C.; Sha ie MS, B.; Sha ie, M.S.; Sulaiman, S.N.; Ib ahim, M.H.; Abdullah SA, C. IoT Wa e Consump ion Moni o -
ing & Ale Sys em. In P oceedings o he 2018 In e na ional Con e ence on Elec ical Enginee ing and In o ma ics (ICELTICs),
Banda Aceh, Indonesia, 19–20 Sep embe 2018. [C ossRe ]
10.
Linaza, M.T.; Posada, J.; Bund, J.; Eise , P.; Qua ulli, M.; Döllne , J.; Pagani, A.; Olaizola, I.G.; Ba iguinha, A.; Moysiadis, T.; e al.
Da a-D i en A i icial In elligence applica ions o sus ainable p ecision ag icul u e. Ag onomy 2021,11, 1227. [C ossRe ]
11.
Shanka , P.; We ne , N.; Selinge , S.; Janssen, O. A i icial In elligence D i en C op P o ec ion Op imiza ion o Sus ainable
Ag icul u e. In P oceedings o he 2020 IEEE/ITU In e na ional Con e ence on A i icial In elligence o Good (AI4G), Gene a,
Swi ze land, 21–25 Sep embe 2020. [C ossRe ]
12.
Naga , H.; Sha ma, R. A Comp ehensi e Su ey on Pes De ec ion Techniques using Image P ocessing. In P oceedings o he
2020 4 h In e na ional Con e ence on In elligen Compu ing and Con ol Sys ems (ICICCS), Madu ai, India, 13–15 May 2020.
[C ossRe ]
13.
Mohan y, S.P.; Hughes, D.; Sala hé, M. Using deep lea ning o Image-Based plan disease de ec ion. F on . Plan Sci.
2016
,7, 1419.
[C ossRe ] [PubMed]
14.
Xin, Z.; Han, L.; Dong, Y.; Shi, Y.; Huang, W.; Han, L.; González-Mo eno, P.; Ma, H.; Ye, H.; Sobeih, T. A Deep Lea ning-Based
app oach o au oma ed yellow us disease de ec ion om High-Resolu ion Hype spec al UAV images. Remo e Sens.
2019
,
11, 1554. [C ossRe ]
15.
T uong, T.; Dinh, A.; Wahid, K.A. An IoT en i onmen al da a collec ion sys em o ungal de ec ion in c op ields. In P oceed-
ings o he 2017 IEEE 30 h Canadian Con e ence on Elec ical and Compu e Enginee ing (CCECE), Windso , ON, Canada,
30 Ap il–3 May 2017. [C ossRe ]
16.
Az a , S.; Nadeem, A.; Ahsan, K.; Mehmood, A.; Siddiqui, M.S.; Saeed, M.; Ash a , M. An IoT-Based sys em o e icien de ec ion
o co on pes . Appl. Sci. 2023,13, 2921. [C ossRe ]
17. Na owBand IOT. A ailable online: h ps://www.3gpp.o g/news-e en s/1733-nio (accessed on 8 Sep embe 2023).
18.
Ins i u o de In es igación y Fo mación Ag a ia y Pesque a (IFAPA). Lis ado de Es aciones. A ailable online: h ps://www.
jun adeandalucia.es/ag icul u aypesca/i apa/ iaweb/web/es aciones (accessed on 4 Augus 2023).
Senso s 2023,23, 9733 19 o 19
19.
Fundación Aquae. ¿Cuál es El País Donde Llue e Más? ¿Y Menos? 2021. A ailable online: h ps://www. undacionaquae.o g/
wiki/en-que-pais-llue e-mas-y-menos-colombia-y-egip o/#:~: ex =%C2%BFCu%C3%A1l%20es%20el%20pa%C3%ADs%20
donde%20m%C3%A1s%20llue e?,mm%20de%20llu ia%20po %20a%C3%B1o (accessed on 11 Oc obe 2023).
20.
Resumen de La E olución de Las P ecipi aciones en España. 2023. A ailable online: h ps://www.aeme .es/documen os/es/
se iciosclima icos/ igilancia_clima/ esumen_p ecipi aciones/ esumen_p ecipi aciones.pd (accessed on 16 Oc obe 2023).
21.
Chen, C.; Huang, Y.; Li, Y.; Chang, C.; Huang, Y. An AIOT based sma ag icul u al sys em o pes s de ec ion. IEEE Access
2020
,
8, 180750–180761. [C ossRe ]
22.
Debauche, O.; Mahmoudi, S.; Elmoula , M.; Mahmoudi, S.A.; Manneback, P.; Lebeau, F. Edge AI-IoT pi o i iga ion, plan
diseases, and pes s iden i ica ion. P ocedia Compu . Sci. 2020,177, 40–48. [C ossRe ]
23.
Ma e ne, N.; Inoue, M. IoT Moni o ing Sys em o Ea ly De ec ion o Ag icul u al Pes s and Diseases. In P oceedings o he 2018
12 h Sou h Eas Asian Technical Uni e si y Conso ium (SEATUC), Yogyaka a, Indonesia, 12–13 Ma ch 2018. [C ossRe ]
24.
Win, T.T.; Ma kon, S. IoT and AI Me hods o Plan Disease De ec ion in Myanma . Mas e ’s Thesis, Kobe Ins i u e o Compu ing,
Hyogo, Japan, 2018. A ailable online: h ps://www. esea chga e.ne /publica ion/326988635_IoT_and_AI_me hods_ o _plan _
disease_de ec ion_in_Myanma (accessed on 4 Sep embe 2023).
Disclaime /Publishe ’s No e:
The s a emen s, opinions and da a con ained in all publica ions a e solely hose o he indi idual
au ho (s) and con ibu o (s) and no o MDPI and/o he edi o (s). MDPI and/o he edi o (s) disclaim esponsibili y o any inju y o
people o p ope y esul ing om any ideas, me hods, ins uc ions o p oduc s e e ed o in he con en .
