sus ainabili y
A icle
Usage o Real Time Machine Vision in Rolling Mill
Jiˇ íDa id *, Pa el Š ec, Ví Paske and Romana Ga zino á


Ci a ion: Da id, J.; Š ec, P.; Paske ,
V.; Ga zino á, R. Usage o Real Time
Machine Vision in Rolling Mill.
Sus ainabili y 2021,13, 3851. h ps://
doi.o g/10.3390/su13073851
Academic Edi o s: Be na di
P anggono, Alex Shen ield and
Augus ine Ikpehai
Recei ed: 1 Feb ua y 2021
Accep ed: 26 Ma ch 2021
Published: 31 Ma ch 2021
Publishe ’s No e: MDPI s ays neu al
wi h ega d o ju isdic ional claims in
published maps and ins i u ional a il-
ia ions.
Copy igh : © 2021 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/).
Depa men o Au oma ion and Compu ing in Indus y, VŠB-Technical Uni e si y Os a a,
17. Lis opadu 2173/15, 70800 Os a a, Czech Republic; [email p o ec ed] (P.Š.); [email p o ec ed] (V.P.);
[email p o ec ed] (R.G.)
*Co espondence: [email p o ec ed]; Tel.: +420-723-858-551
Abs ac :
This a icle deals wi h he issue o compu e ision on a olling mill. The main goal o
his a icle is o desc ibe he designed and implemen ed algo i hm o he au oma ic iden i ica ion
o he cha ac e s ing o bille s on he olling mill. The algo i hm allows he con e sion o image
in o ma ion om he on o he bille , which en e s he olling p ocess, in o a s ing o cha ac e s,
which is u he used o con ol he echnological p ocess. The pu pose o his iden i ica ion is o
p e en he inpu pieces om being con used because di e en pa ame e s o he olling p ocess
a e se o di e en pieces. In sol ing his ask, i was necessa y o design he op imal echnical
equipmen o image cap u e, choose he app op ia e ligh ing, sea ch o ex and ecognize indi idual
symbols, and inse hem in o he con ol sys em. The esea ch me hodology is based on he empi ical-
quan i a i e p inciple, he basis o which is he analysis o expe imen ally ob ained da a (pho og aphs
o bille aces) in eal ope a ing condi ions leading o hei in e p e a ion ( ans o ma ion in o he
shape o a digi al chain). The i s pa o he a icle b ie ly desc ibes he bille iden i ica ion sys em
om he poin o iew o echnology and ha dwa e esou ces. The nex pa s a e de o ed o
he main pa s o he algo i hm o au oma ic iden i ica ion—op ical ecogni ion o s ings and
ecogni ion o indi idual cha ac e s o he chain using a i icial in elligence. The me hod o op ical
cha ac e ecogni ion using a i icial neu al ne wo ks is he basic algo i hm o he sys em o au oma ic
iden i ica ion o bille s and elimina es ambigui ies du ing hei u he p ocessing. Success ul
implemen a ion o he au oma ic inspec ion sys em will inc ease he sha e o ope a ion au oma ion
and lead o ensu ing au oma ic inspec ion o s eel bille s acco ding o he p oduc ion plan. This
issue is ela ed o he end o digi iza ion o indi idual echnological p ocesses in me allu gy and
also o he social sus ainabili y o p ocesses, which means he elimina ion o human e o s in he
managemen o he bille olling p ocess.
Keywo ds:
me allu gy; machine ision; neu al ne wo ks; con ol; applica ions; s eel indus y; op ical
de ices; a i icial in elligence; machine lea ning; in elligen sys ems; a i icial neu al ne wo ks; image
p ocessing; image classi ica ion; eal- ime sys ems
1. In oduc ion
In oday’s indus ial companies and manu ac u ing companies, he emphasis is on
p oduc quali y and in his espec on he numbe o con ol poin s a each s age o
p oduc ion. Checks can be pe o med manually, au oma ically, o in combina ion. Wi h
he g owing mode niza ion o business and he de elopmen o au oma ion echnologies,
many ope a ions a e mo ing o au oma ed managemen . Acco dingly, a ious quali y
con ol me hods a e used a con ol s a ions, such as machine ision, su ace scanning,
p ecision shaping, colo and size measu emen , o inspec ion o semi- inished p oduc s
and p oduc iden i ie s du ing p oduc ion and s o age. Many o hese con ol me hods
use an indus ial came a as an essen ial elemen o p o ide he equi ed in o ma ion in
image o m.
Sus ainabili y 2021,13, 3851. h ps://doi.o g/10.3390/su13073851 h ps://www.mdpi.com/jou nal/sus ainabili y
Sus ainabili y 2021,13, 3851 2 o 18
The applica ion o a i icial in elligence me hods in indus ial applica ions is oday a
end bo h in he ield o p ocess digi iza ion and in he ield o sus ainabili y, whe e i is
also necessa y o apply new me hods and app oaches.
The e has been a lo o alk la ely abou he Indus y 4.0 ini ia i e o he 4 h Indus-
ial Re olu ion. New ends in au oma ion, digi iza ion, o da a exchange in indus ial
p ocesses and echnologies can be hea d om all sides. On he o he hand, he issue can
also be seen in he con ex o sus ainable de elopmen . The abo e-desc ibed solu ion o
using machine ision in a olling mill is an example o a pa ial solu ion con ibu ing o he
digi iza ion o he p ocess in o de o c ea e a so-called “Sma Fac o y”, which combines
he physical and cybe wo ld o p oduc ion, bu also as a solu ion ha is in line wi h
sus ainable de elopmen o socie y in all ac o s—inc easing he quali y o p oduc ion,
educing p oduc ion cos s, and om a social and en i onmen al poin o iew. This con-
nec ion signi ican ly akes he digi iza ion o indus y o a new le el. Bu wha does he
combina ion o he physical and cybe wo ld eally mean o manu ac u ing companies?
Wha oppo uni ies does he digi aliza ion o indus y b ing? The answe can be con ained
in one e m—in elligen da a.
In elligen da a is closely linked o digi iza ion and da a analy ics. We can de ine
hem as da a ha a e unambiguous and cleaned up and ha a e linked di ec ly o a
speci ic ac ion.
In elligen da a is he esul o digi iza ion and da a analy ics. In o de o wo k ou
he in o ma ion and alue ha da a ep esen s wi hin he company, i is i s necessa y o
collec i (in eg a e di e en da a sou ces), p ocess i (e.g., ans o m i in o he same da a
ype o clean i om da a impu i ies) and analyze i (using di e en ypes o da a analysis,
such as p edic i e analysis).
Today, howe e , he ield o in elligen da a aces a la ge numbe o challenges. These
include he speed o new da a, secu i y, and he lack o expe s in he ma ke (wi h
knowledge o he company’s p oduc ion p ocesses and also knowledge o da a analy ics)
o he in eg a ion o se e al di e en da a sou ces. This aises he ques ion o whe he
in elligen da a is eally wo h he e o and wha digi iza ion and da a analy ics ac ually
b ing o manu ac u ing companies. Simply pu , i helps companies unde s and wha and
how i is done, and hen helps iden i y oppo uni ies o imp o emen . In he ield o
digi iza ion, we a e no only alking abou he ans e o he physical wo ld o he cybe
wo ld, he p ocesses o da a collec ion, o he elimina ion o communica ion in pape
o m ( he concep o he “pape less company”). Digi iza ion in he con ex o p oduc ion
also includes concep s such as he digi al win, ad anced da a mining o p oduc ion and
non-p oduc ion p ocesses, o he en i e p oduc ion and p oduc li e cycle.
Digi iza ion is hus a co ne s one o Indus y 4.0. We can say ha i is he echnical
ad ances in digi iza ion ha enable he comple e in e connec ion o he en i e p oduc ion
p ocess and he en i e cus ome -consume chain. Subsequen ly, i is possible o apply
ad anced da a p ocessing me hods, apply s a is ical and p edic ion models, moni o he
beha io o sys ems and machines in eal- ime, o p e en ou ages in p oduc ion p ocesses.
Today, echnological ad ances allow us o collec da a di ec ly om p oduc ion lines
(In e ne o Things: IoT), p e-p ocess i a he local le el and isualize i in clea epo s o
wo k su aces (edge compu ing), send i o he cloud o u he p ocessing (e.g., machine
lea ning), o en ich i wi h non-p oduc ion da a and isualize on mobile de ices.
Indi idual echnologies hus eally become he s anda d and in oday’s highly com-
pe i i e en i onmen a e ce ainly a necessi y o main ain hei own compe i i eness.
Indus y 4.0, he digi iza ion o p oduc ion p ocesses, and in elligen da a a e no a
goal, bu a ool o achie e se goals, iden i y new business models, o op imize p oduc ion
p ocesses. These ools use da a as a basic inpu and p e equisi e o be e and mo e
accu a e decision-making. Digi iza ion and da a analy ics hen ep esen some hing as
a co ne s one o he whole Indus y 4.0 ini ia i e, whe e he physical and cybe ne ic
wo ld is in e connec ed in a sma ac o y in o a sys em ha moni o s p ocesses, c ea es
Sus ainabili y 2021,13, 3851 3 o 18
a digi al win (wo kpiece o p ocess), and is able o make i s own decisions (a i icial
in elligence elemen s).
Applica ions aimed a c ea ing sma ac o ies change he unc ioning no only o
indus y and p oduc ion bu also o socie y as a whole. P oduc i i y is a so e spo in he
Czech labo ma ke . A 3% g ow h in labo p oduc i i y g ow h in 2017 was abo e a e age
a e he economic c isis, bu s ill below p e-c isis le els. The le el o labo p oduc i i y
emains ela i ely low, especially compa ed o de eloped economies, due o signi ican skill
misma ches, he low impac o R&D on inno a ion, and he size o low-skilled indus ies.
To achie e economic alue, he Czech economy mus mo e om low-skilled o high-skilled,
he eby inc easing alue-added and imp o ing he posi ion o companies.
The p ocess o au oma ion and digi iza ion is absolu ely c ucial o he Czech Republic.
Bo h p ocesses lead o a educ ion in p oduc ion cos s, sho ening he p oduc ion cycle,
inc easing p oduc ion e iciency, inc easing he use ul p ope ies o p oduc s and se ices,
humanizing wo k, imp o ing he wo king en i onmen , and inc easing leisu e ime.
Acco ding o a ious s udies and ongoing digi iza ion and au oma ion p ojec s, he
g ea es impac is expec ed on employmen in he middle class, whe e many p o essions
a e ou ine and he e o e mos p one o compu e eplacemen . The au oma ion and
digi iza ion o indus ial indus ies alone may no esul in a educ ion in he numbe o
employees; on he con a y, i may be a way o compensa e o he lack o labo due o
un a o able demog aphic de elopmen s in de eloped coun ies.
Ad anced au oma ion hus jeopa dizes mo e wo k asks han en i e jobs. In line wi h
exis ing indings, i is expec ed ha new echnologies will shi he a io o man-made o
machine-d i en asks, wi h he na u e o he changes a ying bo h in scope and speed
o applica ion.
In e ope abili y, decen aliza ion, in elligence, and econ igu abili y a e ea u es ha
cu en ly b ing o he bene i s o companies, o example in he a eas o sa e y and en i on-
men al s udies. The Indus ial In e ne o Things (IIoT) and he as amoun o da a ha
his ne wo k gene a es c ea e g ea e cla i y in business p ocesses, and hus educe he isk
o human o echnological e o .
Indus ial In e ne o Things (IIoT) echnologies and Sma Indus y sys ems ha e a
di ec impac on imp o ing he sa e y, eliabili y, and sus ainabili y o he p oduc ion en i-
onmen , and ul ima ely on p e en ing nega i e impac s on p oduc quali y and deli e y
ime. G ea e anspa ency also helps o educe was e, leakage, and was e and iden i y
hei o igin, hus minimizing he en i onmen al oo p in o he manu ac u ing plan .
This pape deals wi h he issue o op ical cha ac e ecogni ion using a i icial neu al
ne wo ks. The main pa o he pape is a desc ip ion o he designed and implemen ed
algo i hm o op ical iden i ica ion o a cha ac e s ing om he on o a s eel bille using
a i icial in elligence. The pape has i e sec ions. The in oduc o y chap e ou lines he
issue o p ocess digi iza ion and he de elopmen o sus ainabili y in he cu en condi ions
o p ocess digi iza ion. Sec ion 2o he a icle b ie ly desc ibes he bille iden i ica ion
sys em in e ms o echnological and ha dwa e esou ces. Sec ion 3desc ibes he issue o
op ical chain ecogni ion on s eel bille s. Sec ion 4desc ibes he use o neu al ne wo ks o
he ecogni ion o indi idual elemen s o he chain and he achie ed esul s. The conclusion
o he a icle hen p esen s he achie ed esul s in e ms o he p oposed algo i hm and in
e ms o ope a ional use.
2. Sys em Design o Iden i ica ion o S eel Bille s
The basic semi- inished p oduc s o he p oduc ion o olled wi e a e s eel bille s,
which a e s o ed in he a ea o he olling mill and u he dis ibu ed o he p oduc ion
line acco ding o he equi emen s o he p oduc ion plan. Be o e en e ing he hea ing
u nace, he bille s a e inspec ed only du ing he p ocess in he cha ging ne wo k wi hou
inspec ion om he se ice s a ion. In some cases, inaccu a e checks and e o s in bille
changes occu he e. As he ma e ial p ope ies o indi idual bille s a y g ea ly depending
Sus ainabili y 2021,13, 3851 4 o 18
on hei use o speci ic p oduc s, hei changes esul in a al manu ac u ing de ec s and
la ge inancial losses [1].
In o de o mode nize he ope a ion and limi he exchange o bille s, a bille iden i i-
ca ion sys em based on machine lea ning was designed and subsequen ly implemen ed.
The design o he bille iden i ica ion sys em was ealized by ou basic wo king
p ocedu es di ided acco ding o hema ic asks:
•
The i s basic p ocedu e was he selec ion o indi idual componen s, in he selec ion
o which i was necessa y o obse e he equi emen s and condi ions o ope a ion.
•
The second p ocedu e was o se up and ins all a ne wo k came a, including moun ing
he came a and accesso ies, including wi ing.
•
The hi d p ocedu e was he c ea ion o an algo i hm o adjus he bille con eyo
con ol, including he addi ion o new commands in he p og ammable logic con olle
(PLC) con ol p og am and came a con ol se ings.
•
The ou h p ocedu e was he design o an algo i hm o he iden i ica ion o bille s
using machine lea ning and s o ing he in o ma ion ob ained in he Bille S amping
p og am.
Fu he mo e, he a icle is ocused on he desc ip ion o he p oposed algo i hm o
bille iden i ica ion using machine lea ning [2–4].
Inpu in o ma ion and ope a ing equi emen s
1.
S eel bille s ha e, o iden i ica ion pu poses, a se ial numbe , which is c ea ed by a
ma king au oma ic machine immedia ely a e he bille s exi om he con inuous
cas ing machine.
2.
The nume ic cha ac e con ains a combina ion o 5 o 9 digi s, possibly accompanied
by wo le e s.
3.
The s eel bille s a e a anged on he cha ging g a e and indi idually con eyed o he
hea ing u nace.
4.
Fo he pu pose o bille iden i ica ion, he possibili y o digi izing nume ical cha ac-
e s on he compu e moni o o he ope a o ’s s a ion and he possibili y o a chi ing
he ob ained in o ma ion is equi ed.
The selec ion o echnical means o scanning and subsequen iden i ica ion was
made wi h ega d o he p ope ies o he scanned objec s, which we e s eel bille s. The
dimensions o he bille a e 8 m in leng h, 150 mm in wid h, and 150 mm in heigh . The
weigh o one bille is app oxima ely 2 ons.
The bille s a e a anged on a loading g id and indi idually anspo ed o a hea ing
u nace, om which hey con inue o he olling line a e eaching he equi ed empe a u e.
Fo iden i ica ion pu poses, hey a e ma ked wi h a se ial numbe s ing, which is c ea ed
by an au oma ic s amping machine on he on o he bille . The nume ic cha ac e con ains
a combina ion o 5 o 9 digi s, possibly supplemen ed by wo le e s. The basic componen s
we e an indus ial came a, a came a lens, a came a co e , a came a co e holde , a powe
supply, a ligh e lec o , and ela ed cabling. When choosing an indus ial came a and
accesso ies, i was necessa y o ake in o accoun , in pa icula , he me hod o use and he
p ope ies o he en i onmen . Componen s mus be designed o con inuous ope a ion in
an indus ial en i onmen . Also, he emphasis was on eliable ope a ion, so wa e, and
easonable p ice. Based on hese condi ions, he came a IP 1343 was selec ed by Axis (see
Table 1 o pa ame e s) and lens Compu a TG10Z0513FCS (see Table 2 o pa ame e s).
Sus ainabili y 2021,13, 3851 5 o 18
Table 1. Axis P1343 came a pa ame e s.
Came a Resolu ion 800 ×600
Image senso ype CMOS 1/4” p og essi e scan
Minimum ligh ing alue 0.05 lux
Image comp ession MPEG 4, MJPEG, H.264
Numbe o ames pe second 30
Ope a ing empe a u e −30 ◦C o 50 ◦C
Powe ol age 8–20 V DC
Consump ion 6.4 W
Table 2. Compu a TG10Z0513FCS lens pa ame e s.
Focal Dis ance 5–50 mm
Ape u e 1.3–360
Ape u e con ol Au oma ic DC
Fo ma 1/3”
Type o a achmen CS
Due o he p oduc ion en i onmen ha ing bo h high empe a u es and conside able
dus iness, i was necessa y o place he came a in a came a co e . This co e p o ec s he
came a om dus , mois u e, and mechanical damage. The ype used is TP607H p o ec ion
class IP66, i is designed o a empe a u e ange om
−
30
◦
C o +50
◦
C. The co e was
moun ed on a b acke equipped wi h a swi el join ha allows he came a co e o be
swi eled and il ed o he desi ed posi ion.
Came a powe was p o ided by a linea 12 V/800 mA s abilized powe supply. The
powe supply was placed in he came a housing using a co e ki .
A b acke was made o moun he came a housing, welded om L- ype s eel p o iles
40 mm
×
40 mm wi h dimensions o 300 mm
×
1000 mm, and ea ed wi h g ay pain , see
Figu e 1.
Sus ainabili y 2021, 13, x FOR PEER REVIEW 5 o 17
T
ABLE
2.
Compu a TG10Z0513FCS lens pa ame e s.
Focal Dis ance 5–50 mm
Ape u e 1.3–360
Ape u e con ol Au oma ic DC
Fo ma 1/3“
Type o a achmen CS
Due o he p oduc ion en i onmen ha ing bo h high empe a u es and conside able
dus iness, i was necessa y o place he came a in a came a co e . This co e p o ec s he
came a om dus , mois u e, and mechanical damage. The ype used is TP607H p o ec ion
class IP66, i is designed o a empe a u e ange om −30 °C o +50 °C. The co e was
moun ed on a b acke equipped wi h a swi el join ha allows he came a co e o be
swi eled and il ed o he desi ed posi ion.
Came a powe was p o ided by a linea 12 V/800 mA s abilized powe supply. The
powe supply was placed in he came a housing using a co e ki .
A b acke was made o moun he came a housing, welded om L- ype s eel p o iles
40 mm × 40 mm wi h dimensions o 300 mm × 1000 mm, and ea ed wi h g ay pain , see
Figu e 1.
A halogen e lec o wi h a powe o 100 W was used o illumina e he bille s.
In e connec ion o ne wo k came a and da a ne wo k was sol ed by FTP (Foiled
Twis ed Pai ) (cable ca ego y 5e, cable CYKY-J 3 × 2.5 was used o powe supply o cam-
e a and ligh ing. Exis ing cable ou es we e used o s o e he cabling.
Figu e 1. Came a moun including an incandescen ligh sou ce.
P e-ins alla ion se ings and ins alla ions o ne wo k came as, ne wo k came a se -
ings, lens es ing, and ne wo k communica ion, as well as powe supply unc ions, we e
pe o med. The came a was connec ed o a powe sou ce and a ne wo k cable o connec
i o he compu e . The came a was con igu ed using a web b owse . The ne wo k add ess,
ne mask, and ga eway add ess we e se on he came a. Then he adminis a o passwo d
was se . A e hese s eps, i was necessa y o es a he came a o sa e he newly se
alues. Came a con igu a ion con inued by se ing he ime zone, cu en ime, and ime
synch oniza ion se e ime. The nex s ep was o adjus he image quali y and esolu ion.
The image quali y and esolu ion we e se o maximum, MJPEG was selec ed o he ideo
o ma . The came a was equipped wi h a lens and se e al ocusing we e pe o med, in-
cluding he ex eme posi ions o he lens. This was ollowed by es ing he au o i is unc-
ion by aiming he came a a da k and ligh places al e na ely. In his way, he came a,
lens, and powe supply, i.e., he componen s ha ha e he g ea es impac on he inal
image quali y, we e easily es ed.
The came a, including he lens and powe supply, was placed in he came a housing
o which he hinge b acke is a ached. The loca ion o he came a ins alla ion was chosen
Figu e 1. Came a moun including an incandescen ligh sou ce.
A halogen e lec o wi h a powe o 100 W was used o illumina e he bille s.
In e connec ion o ne wo k came a and da a ne wo k was sol ed by FTP (Foiled
Twis ed Pai ) (cable ca ego y 5e, cable CYKY-J 3
×
2.5 was used o powe supply o
came a and ligh ing. Exis ing cable ou es we e used o s o e he cabling.
P e-ins alla ion se ings and ins alla ions o ne wo k came as, ne wo k came a se -
ings, lens es ing, and ne wo k communica ion, as well as powe supply unc ions, we e
pe o med. The came a was connec ed o a powe sou ce and a ne wo k cable o connec i
o he compu e . The came a was con igu ed using a web b owse . The ne wo k add ess,
ne mask, and ga eway add ess we e se on he came a. Then he adminis a o passwo d
was se . A e hese s eps, i was necessa y o es a he came a o sa e he newly se
alues. Came a con igu a ion con inued by se ing he ime zone, cu en ime, and ime

Sus ainabili y 2021,13, 3851 6 o 18
synch oniza ion se e ime. The nex s ep was o adjus he image quali y and esolu ion.
The image quali y and esolu ion we e se o maximum, MJPEG was selec ed o he ideo
o ma . The came a was equipped wi h a lens and se e al ocusing we e pe o med, includ-
ing he ex eme posi ions o he lens. This was ollowed by es ing he au o i is unc ion by
aiming he came a a da k and ligh places al e na ely. In his way, he came a, lens, and
powe supply, i.e., he componen s ha ha e he g ea es impac on he inal image quali y,
we e easily es ed.
The came a, including he lens and powe supply, was placed in he came a housing
o which he hinge b acke is a ached. The loca ion o he came a ins alla ion was chosen
wi h ega d o he p oduc ion echnology, op ical possibili ies o he came a, cable ou ing,
and came a es ing. A b acke o he came a co e was welded a he selec ed loca ion,
and a ligh e lec o wi h a powe supply e mina ed by a double me al socke wi h IP 55
p o ec ion was placed abo e he console.
The came a was placed in he cen e o he con eyo app oxima ely 30 cm abo e he
passing bille , as shown in Figu e 2. The scanning dis ance wi h espec o he equi ed
de ails and lens capabili ies was de e mined by a came a es app oxima ely one me e
om he came a lens.
Sus ainabili y 2021, 13, x FOR PEER REVIEW 6 o 17
wi h ega d o he p oduc ion echnology, op ical possibili ies o he came a, cable ou -
ing, and came a es ing. A b acke o he came a co e was welded a he selec ed loca-
ion, and a ligh e lec o wi h a powe supply e mina ed by a double me al socke wi h
IP 55 p o ec ion was placed abo e he console.
The came a was placed in he cen e o he con eyo app oxima ely 30 cm abo e he
passing bille , as shown in Figu e 2. The scanning dis ance wi h espec o he equi ed
de ails and lens capabili ies was de e mined by a came a es app oxima ely one me e
om he came a lens.
Figu e 2. Modi ica ion o bille con eyo con ol.
Du ing came a es s, i was ound ha i is no app op ia e o scan bille s in mo ion
(Figu e 3), especially because he e ec o ansmi ed ib a ions, which canno be com-
ple ely elimina ed, signi ican ly impai s he image quali y. Much be e esul s a e
achie ed by scanning bille s a es . Based on hese indings, i was necessa y o s op he
passing bille a a gi en dis ance om he came a be o e sa ing he image.
Figu e 3. A pic u e o a passing bille and a bille a es om he Bille S amping p og am.
The olling o he bille is con olled by a p og ammable logic con olle (PLC) Si-
ma ic S7 300. This PLC con ols he mo emen o semi- inished p oduc s on he cha ge
g a e and con eyo s on which he semi- inished p oduc s a e anspo ed o he hea ing
u nace. The PLC p og am o con eyo con ol had o be modi ied so ha he bille
s opped a he speci ied loca ion, a his poin he equi ed ime emained and con inued
o mo e o he hea ing u nace.
The s a command o he PLC is he “p epa e bille ” command om he main con-
ol sys em. The mo emen o he bille om he illing g a e o he hea ing u nace akes
place in he ollowing s eps, which a e also shown in Figu e 4.
• T ans e o bille om cha ging g id o con eyo No. 1.
• S a con eyo a el 1 and 2.
• S op he con eyo No. 1 and 2 o 5 s a e signaling he senso on con eyo No. 2.
• A e s opping, send he “s op came a” command o he mas e con ol sys em.
Figu e 2. Modi ica ion o bille con eyo con ol.
Du ing came a es s, i was ound ha i is no app op ia e o scan bille s in mo ion
(Figu e 3), especially because he e ec o ansmi ed ib a ions, which canno be com-
ple ely elimina ed, signi ican ly impai s he image quali y. Much be e esul s a e achie ed
by scanning bille s a es . Based on hese indings, i was necessa y o s op he passing
bille a a gi en dis ance om he came a be o e sa ing he image.
Sus ainabili y 2021, 13, x FOR PEER REVIEW 6 o 17
wi h ega d o he p oduc ion echnology, op ical possibili ies o he came a, cable ou -
ing, and came a es ing. A b acke o he came a co e was welded a he selec ed loca-
ion, and a ligh e lec o wi h a powe supply e mina ed by a double me al socke wi h
IP 55 p o ec ion was placed abo e he console.
The came a was placed in he cen e o he con eyo app oxima ely 30 cm abo e he
passing bille , as shown in Figu e 2. The scanning dis ance wi h espec o he equi ed
de ails and lens capabili ies was de e mined by a came a es app oxima ely one me e
om he came a lens.
Figu e 2. Modi ica ion o bille con eyo con ol.
Du ing came a es s, i was ound ha i is no app op ia e o scan bille s in mo ion
(Figu e 3), especially because he e ec o ansmi ed ib a ions, which canno be com-
ple ely elimina ed, signi ican ly impai s he image quali y. Much be e esul s a e
achie ed by scanning bille s a es . Based on hese indings, i was necessa y o s op he
passing bille a a gi en dis ance om he came a be o e sa ing he image.
Figu e 3. A pic u e o a passing bille and a bille a es om he Bille S amping p og am.
The olling o he bille is con olled by a p og ammable logic con olle (PLC) Si-
ma ic S7 300. This PLC con ols he mo emen o semi- inished p oduc s on he cha ge
g a e and con eyo s on which he semi- inished p oduc s a e anspo ed o he hea ing
u nace. The PLC p og am o con eyo con ol had o be modi ied so ha he bille
s opped a he speci ied loca ion, a his poin he equi ed ime emained and con inued
o mo e o he hea ing u nace.
The s a command o he PLC is he “p epa e bille ” command om he main con-
ol sys em. The mo emen o he bille om he illing g a e o he hea ing u nace akes
place in he ollowing s eps, which a e also shown in Figu e 4.
• T ans e o bille om cha ging g id o con eyo No. 1.
• S a con eyo a el 1 and 2.
• S op he con eyo No. 1 and 2 o 5 s a e signaling he senso on con eyo No. 2.
• A e s opping, send he “s op came a” command o he mas e con ol sys em.
Figu e 3. A pic u e o a passing bille and a bille a es om he Bille S amping p og am.
The olling o he bille is con olled by a p og ammable logic con olle (PLC) Sima ic
S7 300. This PLC con ols he mo emen o semi- inished p oduc s on he cha ge g a e
and con eyo s on which he semi- inished p oduc s a e anspo ed o he hea ing u nace.
The PLC p og am o con eyo con ol had o be modi ied so ha he bille s opped a he
Sus ainabili y 2021,13, 3851 7 o 18
speci ied loca ion, a his poin he equi ed ime emained and con inued o mo e o he
hea ing u nace.
The s a command o he PLC is he “p epa e bille ” command om he main con ol
sys em. The mo emen o he bille om he illing g a e o he hea ing u nace akes place
in he ollowing s eps, which a e also shown in Figu e 4.
•T ans e o bille om cha ging g id o con eyo No. 1.
•S a con eyo a el 1 and 2.
•S op he con eyo No. 1 and 2 o 5 s a e signaling he senso on con eyo No. 2.
•A e s opping, send he “s op came a” command o he mas e con ol sys em.
•S a he con eyo a el 1 and 2 a e 5 s.
Sus ainabili y 2021, 13, x FOR PEER REVIEW 7 o 17
• S a he con eyo a el 1 and 2 a e 5 s.
The ollowing new commands ha e been added o he sequence o commands in he
PLC con ol p og am:
• T a el s op a e senso signaling om con eyo No. 2.
• A e s opping, send he “s op came a” command.
• A e 5 s, s a con eyo a el 1 and 2.
All hese changes we e made in STEP 7 e sion 5.5. The bille s op senso was used
by he LT 100L and LR 100L in a ligh ba ie wi h a PA 11B e alua ion uni om Telco.
This solu ion excels abo e all in i s high ib a ion and dus esis ance.
Figu e 4. Sys em unc ion diag am.
Came a con ol and image s o age we e pe o med using he Axis Media Con ol
SDK de elopmen so wa e ( e sion 6.32), which is a ailable on he Axis came a manu-
ac u e ’s websi e. Wi h he help o his so wa e, a p og am was c ea ed o con ol he
unc ions o IP came as. The p og am was un on a came a se e , which was equipped
wi h enough memo y o s o e images. The d i e was c ea ed in Visual S udio 2010 ex-
p ess and C #. This p og amming language c ea ed a p og am wi h a simple g aphical
in e ace called Punching Bille s, see Figu e 5. The g aphical in e ace con ains se e al
unc ion keys ha can be used o call up he espec i e came a and p og am unc ions.
A e s a ing he Bille Punching p og am and displaying he g aphical in e ace, i was
necessa y o se he da a o communica ion wi h he IP came a. These include he IP ad-
d ess o he came a, he login name, and passwo d o he came a adminis a o , he ype
o ideo playback (wi h op ions be ween MJPEG, MPEG 4, and H.264), and he eco ding
pa h. A e p essing he Play Li e bu on, he p og am es ablishes communica ion wi h
he IP came a, e i ies he login name and passwo d, and displays he cu en image om
he IP came a in he p e iew window [5,6].
C ea ing a snapsho was done by calling he Sa e cu en image unc ion in Bille
S amping. The unc ion c ea es an image in he came a and sa es i acco ding o he se
pa h. This unc ion is also a ailable di ec ly in he g aphical in e ace on he SnapSho
bu on. Se e al unc ions ha e been p og ammed using he g aphical in e ace bu ons,
such as he al eady men ioned Play Li e unc ion, he Play File unc ion which plays he
ideo s o ed on he se e , he Reco ding unc ion which s a s eco ding he image om
he came a and sa es he eco ding acco ding o he se pa h, and he S op bu on which
ac i a es he unc ion o s op he cu en ope a ion, i.e., play he cu en image o play he
sa ed ideo. The possibili ies o he p og am a e e y wide and allow you o con ol a
la ge numbe o unc ions and con ol he connec ed came a. The desc ibed bille iden i-
ica ion sys em mainly uses he image sa ing unc ion. O he unc ions a e mainly used
o es he sys em. The Bille Punching p og am can also be con olled using commands
and his op ion is used in he desc ibed image s o age sys em. The came a se e ecei es
Figu e 4. Sys em unc ion diag am.
The ollowing new commands ha e been added o he sequence o commands in he
PLC con ol p og am:
•T a el s op a e senso signaling om con eyo No. 2.
•A e s opping, send he “s op came a” command.
•A e 5 s, s a con eyo a el 1 and 2.
All hese changes we e made in STEP 7 e sion 5.5. The bille s op senso was used by
he LT 100L and LR 100L in a ligh ba ie wi h a PA 11B e alua ion uni om Telco. This
solu ion excels abo e all in i s high ib a ion and dus esis ance.
Came a con ol and image s o age we e pe o med using he Axis Media Con ol SDK
de elopmen so wa e ( e sion 6.32), which is a ailable on he Axis came a manu ac u e ’s
websi e. Wi h he help o his so wa e, a p og am was c ea ed o con ol he unc ions o
IP came as. The p og am was un on a came a se e , which was equipped wi h enough
memo y o s o e images. The d i e was c ea ed in Visual S udio 2010 exp ess and C #.
This p og amming language c ea ed a p og am wi h a simple g aphical in e ace called
Punching Bille s, see Figu e 5. The g aphical in e ace con ains se e al unc ion keys ha
can be used o call up he espec i e came a and p og am unc ions. A e s a ing he
Bille Punching p og am and displaying he g aphical in e ace, i was necessa y o se he
da a o communica ion wi h he IP came a. These include he IP add ess o he came a,
he login name, and passwo d o he came a adminis a o , he ype o ideo playback
(wi h op ions be ween MJPEG, MPEG 4, and H.264), and he eco ding pa h. A e p essing
he Play Li e bu on, he p og am es ablishes communica ion wi h he IP came a, e i ies
he login name and passwo d, and displays he cu en image om he IP came a in he
p e iew window [5,6].
Sus ainabili y 2021,13, 3851 8 o 18
Sus ainabili y 2021, 13, x FOR PEER REVIEW 8 o 17
a command o sa e he image om he pa en con olle and in okes he image s o age
unc ion in Bille Punching o c ea e he image and sa e i acco ding o he se pa h. The
componen o he Sa e Image command is he bille ID ha is used o he image name.
Sa ed images a e accessible h ough he p oduc ion in o ma ion sys em o use s wi h
he app op ia e pe missions. Image a chi ing was se o a leas wo yea s.
Figu e 5. G aphical in e ace o Bille S amping p og am.
3. Tes ing he Op ical Recogni ion o Cha ac e is ics
A pa ial goal o he solu ion was o es he op ical cha ac e ecogni ion. The com-
me cial p og ams Abbyy Fine Reade e sion 12 and F ee image OCR (Op ical Cha ac e
Recogni ion) we e used o his pu pose. These p og ams allow you o con e ex om
a pho o o plain ex o ma . Twen y images om he bille iden i ica ion sys em we e
selec ed o es ing. Selec ed images we e es ed in he men ioned p og ams. All es ed
images we e in JPEG o ma and he p og ams we e se o sa e in x o ma [7].
In he i s phase, he images we e es ed in hei o iginal colo s (Figu e 6), ollowed
by black and whi e images (Figu e 6), and in he inal phase, he images we e u ned in o
nega i e colo s (Figu e 6). Each image was es ed sepa a ely. The es esul s in hese p o-
g ams we e insu icien . Changing he colo spec um also did no p oduce be e esul s.
The p og ams could no ecognize he bille numbe s om he es ed images and also did
no expo sa is ac o ily o he x o ma . The basic p oblem was he high e o a e caused
by he au oma ic ec o iza ion o he OCR e alua ion so wa e. Wi h au oma ic ec o i-
za ion, he so wa e mus be able o de e mine he basic en i ies ha make up an image
based on he colo in o ma ion o each pixel. When applied o embossed cha ac e s on
bille s ha con ained many di e en OCR de o ma ions, he so wa e did no pe o m
well. The in oduc ion o p e-p ocessing ools would imp o e his si ua ion.
Figu e 6. Modi ica ion o images o bille s om he Bille S amping p og am.
3.1. Machine Vision
Compu e ision is one o he mos ad anced a eas o compu e echnology. Used o
ecognize he equi ed in o ma ion om he image cap u ed by he indus ial came a,
compu e ision is widely used, especially in he indus ial sec o , whe e i is e e ed o
as machine ision [8].
Figu e 5. G aphical in e ace o Bille S amping p og am.
C ea ing a snapsho was done by calling he Sa e cu en image unc ion in Bille
S amping. The unc ion c ea es an image in he came a and sa es i acco ding o he se
pa h. This unc ion is also a ailable di ec ly in he g aphical in e ace on he SnapSho
bu on. Se e al unc ions ha e been p og ammed using he g aphical in e ace bu ons,
such as he al eady men ioned Play Li e unc ion, he Play File unc ion which plays he
ideo s o ed on he se e , he Reco ding unc ion which s a s eco ding he image om
he came a and sa es he eco ding acco ding o he se pa h, and he S op bu on which
ac i a es he unc ion o s op he cu en ope a ion, i.e., play he cu en image o play he
sa ed ideo. The possibili ies o he p og am a e e y wide and allow you o con ol a la ge
numbe o unc ions and con ol he connec ed came a. The desc ibed bille iden i ica ion
sys em mainly uses he image sa ing unc ion. O he unc ions a e mainly used o es
he sys em. The Bille Punching p og am can also be con olled using commands and
his op ion is used in he desc ibed image s o age sys em. The came a se e ecei es a
command o sa e he image om he pa en con olle and in okes he image s o age
unc ion in Bille Punching o c ea e he image and sa e i acco ding o he se pa h. The
componen o he Sa e Image command is he bille ID ha is used o he image name.
Sa ed images a e accessible h ough he p oduc ion in o ma ion sys em o use s wi h
he app op ia e pe missions. Image a chi ing was se o a leas wo yea s.
3. Tes ing he Op ical Recogni ion o Cha ac e is ics
A pa ial goal o he solu ion was o es he op ical cha ac e ecogni ion. The com-
me cial p og ams Abbyy Fine Reade e sion 12 and F ee image OCR (Op ical Cha ac e
Recogni ion) we e used o his pu pose. These p og ams allow you o con e ex om
a pho o o plain ex o ma . Twen y images om he bille iden i ica ion sys em we e
selec ed o es ing. Selec ed images we e es ed in he men ioned p og ams. All es ed
images we e in JPEG o ma and he p og ams we e se o sa e in x o ma [7].
In he i s phase, he images we e es ed in hei o iginal colo s (Figu e 6), ollowed
by black and whi e images (Figu e 6), and in he inal phase, he images we e u ned in o
nega i e colo s (Figu e 6). Each image was es ed sepa a ely. The es esul s in hese
p og ams we e insu icien . Changing he colo spec um also did no p oduce be e
esul s. The p og ams could no ecognize he bille numbe s om he es ed images and
also did no expo sa is ac o ily o he x o ma . The basic p oblem was he high e o
a e caused by he au oma ic ec o iza ion o he OCR e alua ion so wa e. Wi h au oma ic
ec o iza ion, he so wa e mus be able o de e mine he basic en i ies ha make up an
image based on he colo in o ma ion o each pixel. When applied o embossed cha ac e s
on bille s ha con ained many di e en OCR de o ma ions, he so wa e did no pe o m
well. The in oduc ion o p e-p ocessing ools would imp o e his si ua ion.
Sus ainabili y 2021,13, 3851 9 o 18
Sus ainabili y 2021, 13, x FOR PEER REVIEW 8 o 17
a command o sa e he image om he pa en con olle and in okes he image s o age
unc ion in Bille Punching o c ea e he image and sa e i acco ding o he se pa h. The
componen o he Sa e Image command is he bille ID ha is used o he image name.
Sa ed images a e accessible h ough he p oduc ion in o ma ion sys em o use s wi h
he app op ia e pe missions. Image a chi ing was se o a leas wo yea s.
Figu e 5. G aphical in e ace o Bille S amping p og am.
3. Tes ing he Op ical Recogni ion o Cha ac e is ics
A pa ial goal o he solu ion was o es he op ical cha ac e ecogni ion. The com-
me cial p og ams Abbyy Fine Reade e sion 12 and F ee image OCR (Op ical Cha ac e
Recogni ion) we e used o his pu pose. These p og ams allow you o con e ex om
a pho o o plain ex o ma . Twen y images om he bille iden i ica ion sys em we e
selec ed o es ing. Selec ed images we e es ed in he men ioned p og ams. All es ed
images we e in JPEG o ma and he p og ams we e se o sa e in x o ma [7].
In he i s phase, he images we e es ed in hei o iginal colo s (Figu e 6), ollowed
by black and whi e images (Figu e 6), and in he inal phase, he images we e u ned in o
nega i e colo s (Figu e 6). Each image was es ed sepa a ely. The es esul s in hese p o-
g ams we e insu icien . Changing he colo spec um also did no p oduce be e esul s.
The p og ams could no ecognize he bille numbe s om he es ed images and also did
no expo sa is ac o ily o he x o ma . The basic p oblem was he high e o a e caused
by he au oma ic ec o iza ion o he OCR e alua ion so wa e. Wi h au oma ic ec o i-
za ion, he so wa e mus be able o de e mine he basic en i ies ha make up an image
based on he colo in o ma ion o each pixel. When applied o embossed cha ac e s on
bille s ha con ained many di e en OCR de o ma ions, he so wa e did no pe o m
well. The in oduc ion o p e-p ocessing ools would imp o e his si ua ion.
Figu e 6. Modi ica ion o images o bille s om he Bille S amping p og am.
3.1. Machine Vision
Compu e ision is one o he mos ad anced a eas o compu e echnology. Used o
ecognize he equi ed in o ma ion om he image cap u ed by he indus ial came a,
compu e ision is widely used, especially in he indus ial sec o , whe e i is e e ed o
as machine ision [8].
Figu e 6. Modi ica ion o images o bille s om he Bille S amping p og am.
3.1. Machine Vision
Compu e ision is one o he mos ad anced a eas o compu e echnology. Used
o ecognize he equi ed in o ma ion om he image cap u ed by he indus ial came a,
compu e ision is widely used, especially in he indus ial sec o , whe e i is e e ed o as
machine ision [8].
Compu e ision basically eplaces human abili ies in e ms o unde s anding and
in e p e a ion o he image by echnical means. Typical asks in machine ision may be he
ecogni ion and coun ing o p oduc s using indus ial came as, posi ioning, dimensioning,
o op ical quali y con ol. In conjunc ion wi h o he p oduc handling sys ems, some
manu ac u ing ope a ions can be ully au oma ed using machine ision sys ems.
The basis o a eliable machine ision sys em is sui able illumina ion o he scanned
scene in which he scanned objec s a e loca ed [
9
]. The image da a ob ained by a scanning
de ice, such as a digi al indus ial came a, is hen p ocessed by he sys em o ecognize
he desi ed in o ma ion. The sys em is connec ed o o he con ol pa s o he p oduc ion
line, as equi ed, wi h which i communica es and can au onomously in e ene in he
p oduc ion p ocess o p o ide signaling o he ope a o . F om au oma ed in e en ions,
we can conside s opping he machine, disca ding de ec i e and poo quali y p oduc s,
o a ing pa s in inapp op ia e posi ions, e c. Access o he sys em can also be enabled
emo ely ia an applica ion o web in e ace.
Machine lea ning me hods ha e been used o imp o e compu e ision o iden i y
embossed cha ac e s. The e a e many such me hods, o example [
10
–
15
], which ha e
ce ain limi a ions. In [
16
] o -line cha ac e ecogni ion is p esen ed, which is applied
o he bille iden i ica ion sys em in he s eel indus y. Iden i ica ion cha ac e s a e so
damaged in his applica ion ha we need noise- esis an me hods and obus cha ac e
segmen a ion and ex ac ion algo i hms. We p opose new local me hods o adap i e
h esholding and cha ac e ex ac ion. We use a subspace classi ie using KLT (Ka hunen–
Loè e T ans o m—p incipal componen analysis— obus pa e n ecogni ion me hod) o
classi y cha ac e s. The me hods ha e been es ed on eal indus ial applica ions and ha e
p o en o be success ul and allow implemen a ion in echnological p ocesses.
Unlike he abo e publica ions, he p oposed machine lea ning solu ion o he iden i-
ica ion o embossed cha ac e s is unique o he me allu gical indus y and wo ks online
in eal- ime. The basic di e ence is in he me hod o inding cha ac e s and hus in he
iden i ica ion o he whole s ing a he head o he bille and in he me hod o lea ning
he cha ac e , whe e only selec ed pixels a e used o lea ning and he e o e no he whole
pa e n. This app oach allows he chosen p inciple o be used o suppo eal- ime play and
o con ol models ha mus be simple, as , bu also accu a e. Ano he di e ence is ha
e en hough no il e is applied in he me hod when digi izing he bille ace, he me hod is
comple ely accu a e.
3.2. Techniques o Reduce Cha ac e Reading E o
An image may con ain (and usually does) con ain ce ain e o s, whe he i is he
inpu de ice (came a), he su ace o he ma e ial, o he impe ec ions o he cha ac e s
hemsel es. Some o he main e o s and ways o elimina e hem a e [17]:
•
Remo al o s ains—This is a basic p ocess o educing e o a es. Spo s a e usually
caused by he di y su ace o he senso (came a). The mos e ec i e me hod o s op
his e o is p e en ion, by egula ly cleaning he came a lens and gene al ha dwa e
scanning. Once his e o is embedded in he esul ing image om which he cha -
Sus ainabili y 2021,13, 3851 16 o 18
Al hough he sys em enables au o-associa i e p ocessing o embossed numbe s, si -
ua ions o misiden i ica ion occu (o he o de o 0.01%), which a e, howe e , caused by
he non-embossing o he symbol in he iden i ica ion chain and hus by he e o o he
p e ious echnological p ocess.
5. Conclusions
This a icle p esen s a pa ial pa o he machine ision sys em o au oma ic iden-
i ica ion o he cha ac e s ing o bille s on he olling mill line. The sys em enables he
con e sion o image in o ma ion om he on side o he bille in o a cha ac e s ing,
which is u he used o con ol he echnological p ocess o olling and o a chi e he
p ocess pa ame e s. The iden i ied cha ac e s ing he e se es as a unique iden i ie o he
bille . Cu en ly, all pa s o his con ol sys em a e in ull ope a ing mode.
The c ea ion o his sys em was s a ed on he basis o ope a ional equi emen s, due
o highe au oma ion o ope a ion and ensu ing he con ol o s eel bille s acco ding o he
p oduc ion plan. The machine ision sys em uses a came a o scan he a ea in on o he
u nace and allows au oma ic inspec ion. The ou pu o he desc ibed pa o he sys em
is digi al in o ma ion o nume ical cha ac e s on he compu e moni o o he ope a o ’s
s a ion and enabling a chi ing o he ob ained in o ma ion. This educes he ope a o ’s
wo kload and educes manu ac u ing e o s. The iden i ied cha ac e s ing is he esul o
a sys em o an algo i hm o au oma ically iden i ying he cha ac e s ing and indi idual
cha ac e s and an algo i hm o au oma ically digi izing indi idual cha ac e s.
Heu is ic algo i hms o inding he iden i ica ion chain in combina ion wi h mul i-
laye a i icial neu al ne wo ks o he pe cep on ype wi h opology 69-38-42-10 we e used
o au oma ic iden i ica ion o embossed numbe s.
A aining se o pho og aphs o he on s o bille s wi h iden i ica ion s ings was
c ea ed o he lea ning p ocess. The iden i ica ion s ings we e di ided in o indi idual
cha ac e s, hus c ea ing a aining se o he iden i ica ion o indi idual cha ac e s. The
aining se o his es ing con ained 45 pa e ns in he o iginal, black and whi e, and
nega i e colo s wi h nume ical cha ac e s on he bille s. The applica ion o a i icial
neu al ne wo ks in he ield o image p ocessing is no new, bu he uniqueness o he
p oposed solu ion lies in he modi ica ion o his p ocessing, which esul s in a signi ican
simpli ica ion. The designed and implemen ed algo i hm o au oma ic iden i ica ion
o embossed numbe s does no supp ess backg ound noise, so i does no pe o m p e-
p ocessing and segmen a ion, which g ea ly simpli ies he solu ion and speeds up he
inal p ocessing while main aining 100% classi ica ion o embossed numbe s. These side
e ec s a e elimina ed by he p oposed neu al ne wo k. The p oposed sys em o au oma ic
iden i ica ion does no educe he p oduc ion cycle o echnological p ocesses, as he
p ocess is no con inuous and p o ides enough ime (app oxima ely 2 min) o he en i e
p ocessing. Howe e , he p oposed algo i hm p o ides digi al in o ma ion in a ma e o
seconds and is he e o e usable in con inuous p ocesses.
Al hough he sys em enables au o-associa i e p ocessing o embossed numbe s, si -
ua ions o misiden i ica ion occu (o he o de o 0.01%), which a e, howe e , caused
by he non-embossing o he symbol in he iden i ica ion chain and hus by he e o o
he p e ious echnological p ocess. The whole solu ion was done on eal pic u es. This
solu ion is an example o he digi iza ion o echnological p ocesses, which is e lec ed in
sus ainabili y in he social and economic ield. This example o a solu ion also shows he
possibili ies o a i icial neu al ne wo ks and o he me hods o a i icial in elligence, in
gene al, o sol e complex p oblems in indus ial and non-indus ial en i onmen s. The
p oposed solu ion is unique o he ollowing easons:
- i is applied in a me allu gical en i onmen , which shows la ge di e ences in digi al
image p ocessing (impu i ies, ligh ing, he e ogenei y o shapes and su aces, speci ic
empe a u e condi ions, e c.);
- he me hod can ind he cha ac e s ing on he ace o he bille ;
-
he me hod can iden i y a cha ac e wi hou he use o a il e in digi al image p ocessing;

Sus ainabili y 2021,13, 3851 17 o 18
-
he me hod iden i ies he cha ac e , while only selec ed pixels o he cha ac e a e
used o lea ning, which enables eal- ime wo k;
-
he me hod is accu a e and as , which allows i o be applied o decision suppo ,
p ocess managemen , o quali y con ol sys ems;
- he me hod wo ks using s anda d ha dwa e and so wa e.
This solu ion is an example o an inno a i e app oach o he issue o digi iza ion o
echnological p ocesses, which is e lec ed in sus ainabili y in he social and economic ield.
Au ho Con ibu ions:
Concep ualiza ion, J.D., P.Š., V.P. and R.G.; me hodology, J.D., P.Š., V.P. and
R.G.; so wa e, V.P.; alida ion, J.D., P.Š. and V.P.; o mal analysis, R.G.; in es iga ion, J.D. and P.Š.;
esou ces, V.P. and R.G.; da a cu a ion, J.D. and V.P.; w i ing—o iginal d a p epa a ion, J.D., P.Š.,
V.P. and R.G.; w i ing— e iew and edi ing, J.D. and P.Š.; isualiza ion, J.D., P.Š. and V.P.; supe ision,
J.D., P.Š., V.P. and R.G.; p ojec adminis a ion, J.D. and P.Š.; unding acquisi ion, J.D. All au ho s
ha e ead and ag eed o he published e sion o he manusc ip .
Funding:
The wo k was suppo ed by he speci ic uni e si y esea ch o Minis y o Educa ion,
You h and Spo s o he Czech Republic No. SP2020/18, SP2020/61 and SP2021/23.
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 : No applicable.
Con lic s o In e es : The au ho s decla e no con lic o in e es .
Re e ences
1.
Da id, J.; Š ec, P.; F ische , R. Suppo o main enance and echnology con ol on slab de ice o con inuous cas ing. In
P oceedings o he Me al 2013: 22nd In e na ional Con e ence on Me allu gy and Ma e ials, B no, Czech Republic, 15–17 May
2013; pp. 1650–1655.
2.
Šonka, M.; Hla áˇc, V.; Boyle, R. Image p ocessing, Analysis, and Machine Vision, 3 d ed.; Thomson: To on o, ON, USA, 2008; p. 829.
3.
Ho ák, K.; Kalo á, I.; Pe yo ský, P.; Rich e , M. Poˇcí aˇco éVidˇení, 1s ed.; VysokéUˇceníTechnické B nˇe: B no, Czech Republic,
2008; p. 132.
4. Bo ik, A. Handbook o Image and Video P ocessing, 2nd ed.; Academic P ess: London, UK, 2000; p. 891.
5. Szeliski, R. Compu e Vision: Algo i hms and Applica ions, 1s ed.; Sp inge Publishing: New Yo k, NY, USA, 2011; p. 812.
6. Fo sy h, D.A.; Ponce, J. Compu e Vision: A Mode n App oach, 2nd ed.; Pea son: London, UK, 2011; p. 800.
7.
Seidl, D.; Koš ial, P.; Janˇcíko á, Z.; Ružiak, I.; Rusnáko á, S.; Fa kašo á, M. Modal analysis—Measu emen s e sus FEM and
a i icial neu al ne wo ks simula ion. Commun. Compu . In . Sci. 2011,188, 170–175.
8.
Puja i, P.; Ka im, M.R.; Sewak, M. P ac ical Con olu ional Neu al Ne wo ks, 2nd ed.; Pack Publishing—ebooks Accoun : Bi ming-
ham, UK, 2018; p. 218.
9.
Janˇcíko á, Z.; Koš ial, P.; Bakošo á, D.; Seidl, D.; Da id, J.; Valíˇcek, J.; Ha niˇcá o á, M. Neu al Model o P edic ion o Ti es
Eigen equencies. Ad . S uc . Ma e . 2013,41, 1–11.
10.
Rui, W.; Dong, A. The Me hod o Pa e n Recogni ion based on Weigh ed In ui ionis ic Fuzzy Rela i e En opy. In P oceedings o
he 2013 Fou h In e na ional Con e ence on Digi al Manu ac u ing and Au oma ion (ICDMA), Qingdao, China, 29–30 June 2013;
pp. 1487–1489.
11.
Yang, P.; Xing, K.; Huang, J. A No el Fea u e Reduc ion Me hod o Real-Time EMG Pa e n Recogni ion Sys em. In P oceedings
o he 2013 25 h Chinese Con ol and Decision Con e ence (CCDC) 2013, Guiyang, China, 25–27 May 2013; pp. 1500–1505.
12. Zhang, J.; Liang, L. New Pa e n Recogni ion Me hod based on Wa ele De-Noising and Ke nel P incipal Componen Analysis.
Appl. Mech. Ma e . 2012,235, 74–82. [C ossRe ]
13.
Yuan, X.; Wu, X.; Wei, X. A Pa e n Recogni ion Me hod Using Pa allel Py amid Neu al Ne wo ks. In P oceedings o he 2011
In e na ional Con e ence on Compu e s, Communica ions, Con ol and Au oma ion (CCCA 2011), Hokkaido, Japan, 1–2 Feb ua y
2011; Volume I, pp. 284–287.
14.
Valdez, F.; Melin, P.; Cas illo, O. Fuzzy Logic Adap a ion o a Hyb id E olu iona y Me hod o Pa e n Recogni ion. In
P oceedings o he Join 2009 In e na ional Fuzzy Sys ems Associa ion Wo ld Cong ess and 2009 eu opean socie y o Fuzzy Logic
and Technology Con e ence 2009, Lisbon, Po ugal, 20–24 July 2009; pp. 815–820.
15.
Yu, J.; Zheng, X.; Wang, S. A deep au oencode ea u e lea ning me hod o p ocess pa e n ecogni ion. J. P ocess Con ol
2019
,79,
1–15. [C ossRe ]
16. Kandel, E.R. P inciples o Neu al Science, 5 h ed.; McG aw-Hill: New Yo k, NY, USA, 2012; p. 1504.
17. Theodo idis, S.; Kou oumbas, K. Pa e n Recogni ion, 4 h ed.; Academic P ess: Camb idge, MA, USA, 2008.
Sus ainabili y 2021,13, 3851 18 o 18
18.
Rice, S.V.; Nagy, G.; Na ke , T.A. Op ical Cha ac e Recogni ion: An Illus a ed Guide o he F on ie ; Sp inge : New Yo k, NY,
USA, 1999.
19.
Da id, J.; F ische , R.; S ana o a, M. Usage o RFID wi eless iden i ica ion echnology o suppo decision making in s eel wo ks.
In P oceedings o he Me al 2012: 21s In e na ional Con e ence on Me allu gy and Ma e ials, B no, Czech Republic, 23–25 May
2012; pp. 1734–1738.
20.
F ische , R.; Da id, J.; S ec, P.; K ejca , O. Usage o analy ical diagnos ics when e alua ing unc ional su ace ma e ial de ec s.
Me alu gija 2015,54, 667–670.