361
In e na ional Jou nal o Ad ance and Applied Resea ch
www.ijaa .co.in
ISSN – 2347-7075
Impac Fac o – 8.141
Pee Re iewed
Bi-Mon hly
Vol. 6 No. 38
Sep embe - Oc obe - 2025
Applica ion o Deep Lea ning echnique in Image Mo phology
M . Akshay Jag ap
Assis an P o esso ,
D . D. Y. Pa il A s, Comme ce and Science College Aku di, Pune, Maha ash a.
Co esponding Au ho – M . Akshay Jag ap
DOI - 10.5281/zenodo.17317547
Abs ac :
Coun ing ehicles along he busy oad o Pune plays an impo an ole in making he
decision o a ic con ol and managemen pu poses. Nowadays, an in elligen ehicle managemen
sys em uses mode n AI deep lea ning and Image p ocessing echniques. The pape discussed he
implemen a ion o ehicle de ec ion algo i hms using image mo phological ope a ion. This p ocess
emo es he noise p esen in he cap u ed image da a and de ec s he mo ing ehicle. Con inuous
acking he de ec ed ehicle along he segmen ed oad helps o coun he numbe o ehicles.
Keywo ds: Vehicle de ec ion, an In elligen Vehicula Managemen sys em, ehicle coun , Image
P ocessing
Abb e ia ions: CCTV, came a; HD, High De ini ion; JM Road, Jangali Maha aj Road.
In oduc ion:
In ecen yea s, he popula ion o Pune
ci y has inc eased. In 2020, he popula ion o
Pune ci y has been eaching 6,62,900 which is
2.7 pe cen han p e ious yea ‟s popula ion
(2019). Simila ly, he e is a signi ican
inc ease in he ehicles ha a e mo ing in he
cen al pa s o Pune ci y. As pe he s a is ics
published by he Pune Regional T a ic
Depa men (MH-12), he e a e nea ly 3.62
million egis e ed ehicles in he ci y.
The e o e, he ci y su e s hea y a ic
conges ion ac oss majo loca ions in he ci y.
T a ic conges ion occu s due o
limi ed oad in as uc u e such as an inc ease
in he numbe o ehicles on oads and poo
oad de elopmen . I is no possible o ul ill
he demand o cons uc ing new oads wi h he
limi ed a ailabili y o he land. An In elligen
ehicula managemen sys em consis s o
a ious applica ions which a e used o p ocess,
analyze and imp o e he a ic managemen
sys em.
Digi al image p ocessing echniques
play a c ucial ole in eal ime applica ions
such as objec de ec ion in ehicle de ec ion
sys em [1,3]. Some o he impo an
applica ions using image p ocessing
echniques include iden i ica ion and coun ing
o he numbe o ehicles mo ing along he
oad, Numbe o ehicles pa ked in pa king
spaces and many mo e. The ehicle de ec ion
echnique helps o design sys ema ic ehicle
pa king spaces and o de e mine he
conges ion le els in he ci y.
Au oma ic ehicle de ec ion and
coun ing o ehicles equi es images o ideo
da a gene a ed om he CCTVs and came as
ins alled on oads which a e used o moni o
he a ic low [2]. Image p ocessing
algo i hms can pe o m he ma hema ical
manipula ions on hese eco ded da a and
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
M . Akshay Jag ap
362
p o ide he de ec ion and coun ing he numbe
o ehicles p esen along he oad.
Image Mo phology In Vehicle De ec ion:
Mo phology is an image p ocessing
echnique ha has non-linea ope a ions which
a e ela ed o mo phological ea u es o shapes
o an image.
This echnique wo ks on he ela i e
o de o he pixel alues and hus i is bes
sui ed o bina y image p ocessing. The
Mo phology ope a ion uses se heo y [5, 6,
and 8]. In he bina y mo phological p ocess,
an image does assume o ha e only wo
in ensi y pixels ha a e whi e and black. The
whi e pixel ep esen s he o eg ound po ion
o an image and black ep esen s he
backg ound o an image.
The ounda ion o he Mo phological
p ocess is no hing bu he geome y such as
ci cles, sphe e, ui s, ehicles, e c. which is
p esen in an image [2]. I is necessa y o ha e
p obes wi h some shape and size. These
p obes ha e a ma ix called “s uc u ing
elemen ” ha ecognized he pixel in an
image. This also ecognizes nea by pixels
p esen in an image. I is hen posi ioned o all
possible loca ions in an image. I is also
compa ed wi h nea by pixels.
Mo phological ope a ion has wo
basic componen s which a e called as
“Dila ion” and “E osion”. Dila ion e e s o
expansion and e osion e e s o comp ession.
The s uc u ing elemen de e mines he ex en
o expansion and comp ession.
A. Dila ion and E osion in Mo phology:
Dila ion p ocess is p ima ily used o
expand o widen he shape o an objec .
S uc u ing elemen con ols he size o he
shape. I is also used o ex ac and elimina e
ea u e ec o s om an image.
In his ope a ion, he image ea u es
a e conside ed as ea u e ec o s. These
ec o s combine and mo e h oughou he
image. Le “A” and “B” be he se in “N‟”
space wi h „i‟ and „j‟ be i s co esponding
elemen s.
The ec o s 𝑖 = (𝑖1, …. 𝑖𝑁) and 𝑗 =
(𝑗1, …. 𝑗𝑁) be he N- uples o co-o dina es.
These co-o dina es a e used o speci y he
loca ions o o eg ound pixels o an image.
The s uc u ing elemen “B” dila es an
image “A”. These se s o a ailable ec o s
combine all elemen pai s. The dila ion
ope a ion is gi en as 𝐴∅𝐵 = {𝑏 ∈ 𝑁|𝑏 = 𝑖 + 𝑗
o some 𝑖 ∈ 𝐴 and𝑗 ∈ 𝐵.
E osion is exac ly e e se ope a ion o
dila ion. This p ocess s ips down he
bounda ies o shapes in images. E osion also
ac s like a linea LPF which is used o emo e
he noise i p esen . The assigned elemen used
o he educ ion and p obing he objec ‟s
shapes a e p esen in images. E osion is
modeled by se heo y ope a ion. The p ocess
is gi en as
B. Edge De ec ion in Mo phology based
ehicle De ec ion:
The edge de ec ion [3] is a echnique
whe e he images a e segmen ed in o a ying
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
M . Akshay Jag ap
363
g ey le els. The edges in an image a e he
g oupings o connec ed pixels which esul in
he bounda ies o an objec [8, 10].
Applica ions whe e he edge de ec ion
p inciple used a e pa e n ecogni ion,
Mo phology and ea u e ex ac ion.
The e a e wo ypes o edge de ec ion
ope a o s: G adien [14] and Gaussian.
G adien based Edge De ec o
Ope a o :
This ope a o compu es he i s o de
de i a i es in an image. P ewi , Sobel and
Robe ope a o a e some o he a ailable edge
de ec o ope a o s.
A. Sobel ope a o is also known as as and
e icien disc e e di e en ia o [5, 6, 7]. I is
used o compu e he g adien app oxima ion
o he in ensi y le els o an image in edge
de ec ion. A he compu a ion s age, he sobel
ope a o p oduces a g adien ec o which hen
con ol ed wi h he inpu image.
Gene ally, Sobel Ope a o is used o
measu emen o 2-D spa ial g adien , and i
wo ks on high equency egions ha consis
o edge o an objec in he image. This
ope a o has a pai o 3x3 con olu ional ma ix
ke nels. This pai includes one basic ke nel,
and one is i s 90-deg ee o a ional e sion as
shown in igu e 3.
Con olu ion is pe o med using hese
ke nels ha gene a e he edges [8]. The
con olu ion is pe o med in he ho izon al and
e ical di ec ion. In his p ocess, oin nel
mo es along he ho izon al di ec ion, and i s
90-deg ee coun e pa mo es along he e ical
di ec ion. A ho izon al and e ical ke nel
gene a es wo g adien componen s. The
combina ions o hese wo g adien s a e used
o ind:
An absolu e magni ude o he g adien o
each pixel poin s.
P o ides o ien a ion o ha g adien .
Le 𝐻𝑥 and 𝐻𝑦 be he g adien
calcula ed along “x” and “y” di ec ion.
Following equa ions gi es he magni ude o
he g adien .
The equa ion o he o a ional angle o he
g adien is calcula ed as (3).
B. P ewi ope a o is used o de ec magni ude
and o ien a ion o an image [16]. I gene ally
de ec s ho izon al and e ical edges p esen in
an image. Such Ope a o is mainly used o
p oduce he sum o he squa e o di e ences
be ween nea by pixels wi h he help o a
disc e e di e en ia o .
Gaussian based Edge De ec o : This
ope a o compu es second o de de i a i es in
an image. These
ope a o s a e he canny edge de ec o and
Laplacian o Gaussian ope a o .
De ec ion using he canny edge de ec o is
he mul i-s age p ocess. This p ocess has
he ollowing s eps: Gaussian il e ke nel
is used o he noise emo al. Size o his
ke nel depends on he blu ing e ec . The
equa ion o The Gaussian il e ke nel o
he size (2m+1) X (2m+1) is gi en in
equa ion (4)
Whe e 𝑖 ≤ 𝑗, 𝑗 ≤ (2𝑚 + 1)
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
M . Akshay Jag ap
364
G adien calcula ion: G adien de ec ion
is used o measu ing in ensi ies o an
objec edges and he di ec ion o he
g adien o an image.
Non-maximum Supp ession The edge
de ec ion algo i hm pe o ms he
calcula ion o g adien in ensi y and
p ocess mo es owa d he pixel poin s
ha ha e maximum alues o g adien .
Double Th eshold: I is used o ind
s ong, weak and non- ele an pixels in
an image. In his high h eshold is used o
ind he s ong pixel and Low h eshold is
used o calcula e he weak pixels.
Laplacian o Gaussian is used o
loca e edge poin s. This ope a o uses Laplace
o he second de i a i e o an image. A canny
edge de ec o is used o ex ac he ea u e in
an image. This ope a o de ec s he edges o
objec s based on low e o a e, accu a ely
localized edge poin s in he images.
Objec ecogni ion using a ious edge
de ec ion echniques has been published in
pape [9]. An image is segmen ed using sobel,
O su [8], Canny and Gene ic algo i hm [5,7].
Resul shows ha he objec bounda ies shape
has been gene a ed om sobel edge de ec o
han gene ic algo i hm and O su algo i hm.
P oposed Me hodology:
The objec i e o his p oposed
me hodology is o iden i y and coun he
numbe o ehicles mo ing along he JM oad
Pune India. The JM Road is he p ime loca ion
in Pune ci y and has obse ed he daily hea y
a ic conges ion in business hou s.
To educe he a ic conges ion le el
and o u ilize he a ailable pa king spaces
along his oad, an in elligen ehicle
managemen sys em is needed.
To s udy he amoun o low o
ehicles on his oad, image p ocessing-based
mo phology echnique has been implemen ed.
Fo his pu pose, ehicles da a a e needed.
Quali y image sensing de ices a e equi ed o
collec his da a. The quali y o his da a
depends on he ype o sensing de ices, ligh ,
illumina ion and wea he condi ions. CCTV
and HD came as a e moun ed on he oad o
eco d image and ideo da a o mo ing
ehicles.
In a ehicle de ec ion p ocess, image
sensing de ice posi ioning plays a c ucial
ole. Image sensing de ices like ideo
eco de s o came as a e used. This ideo
came a is used o eco ding ehicle da a
om he oad a speci ic ime du a ion.
The eco ded ideo is hen b oken in o
he numbe o ames a speci ic ime
in e als. These ames a e hen
indi idually gi en o an image
enhancemen s ep whe e he p ocess o
noise educ ion is ca ied ou .
The eco ded ideo ames a e hen
con e ed in o g ey le els. The bina y
o ms o hese images a e ob ained using
h esholding unc ion. Se ies o
h esholding ope a ions a e pe o med o
gene a e mul iple bina y images.
In edge de ec ion s ep, sobel edge
ope a o is used o p oduce edges. Se ies
o mo phological dila ion is pe o med in
ho izon al, e ical and 45-deg ee
di ec ions. The esul s ob ained om
sobel ope a o a e shown in igu e 7.
Bina y illing p ocess: Filling o holes
esul s in he sepa a ion o backg ound
om an image. The o eg ound image
p esen in an image is like shape o an
objec .
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
M . Akshay Jag ap
365
Resul s:
Table 1. Vehicle Coun a Time In e al
Conclusion:
In elligen Vehicle managemen
sys em consis s o a ious echnological
design implemen a ions. Some o he
implemen a ions a e based on mo phological
image p ocessing echniques. Accu a e
coun ing o he numbe o ehicles ha a e
mo ing along he JM oad in Pune ci y is
achie ed using objec de ec ion algo i hm
using mo phological image p ocessing.
Fu u e Scope:
The ehicle de ec ion and coun ing he
numbe o ehicles mo ing on he oad using
deep lea ning is an impo an applica ion. This
can be used in he sma ci y de elopmen
sys em o analyzing a ic conges ion and o
imp o e he in a ci y anspo a ion sys em.
Fu u e wo k should concen a e mainly on
be e e iciency and imp o emen in he
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
M . Akshay Jag ap
366
p oposed echnique. The echnique should use
o he a ailable edge de ec o s like canny,
laplacian edge de ec o s. Resul s o each
should be compa able and e ec i ely used o
implemen objec de ec ion sys ems. Bes
quali y High esolu ion came a mus be used
o collec ideo da a. The p oposed
implemen a ion can be ex ended o ehicle
acking pu pose on he highways.
Acknowledgemen :
The au ho s would like o exp ess
g a e ulness owa ds D . Mohan Waman si
P incipal, D . D. Y. Pa il A s, Comme ce and
Science College Aku di Pune, India o
suppo s.
Re e ences:
1. Bhaska P em Kuma , Sue -Peng Yong;
Image P ocessing Based Vehicle
De ec ion and T acking Me hod; 978-
14799-0059-6/13/2014 IEEE
2. Ajmal Aisha, Hussain Ib ahim M.;
Vehicle De ec ion Using Mo phological
Image P ocessing Technique; MCIT 2010
3. Kau Jasp ee ; Implemen a ion o Sma
Pa king using A i icial In elligence;
In e na ional Jou nal o Scien i ic
De elopmen and Resea ch (IJSDR);
ISSN 2455-2631, Augus 2019
4. Cai Bill Yang, Al a ez Rica do, Si
Michelle,Dua e Fabio and Ra i Ca lo;
Deep Lea ning Based Video Sys em o
Accu a e and Real-Time Pa king
Measu emen ; IEEE In e ne o Things
Jou nal: Special Issue on Enabling a
Sma Ci y: In e ne o hings me AI, 20
Feb 2019
5. P u ha Y M, Anu adha S G;
Mo phological Image P ocessing
App oach o Vehicle De ec ion o Real-
Time T a ic Analysis; In e na ional
Jou nal o Enginee ing Resea ch and
Technology (IJERT), ISSN: 2278-0181,
Vol. 3 Issue 5, May - 2014.
6. F ıas-Velazquez And es, Jo ge Nino-
Cas aneda Jo ge Oswaldo, Jelaca Ved an,
Pizu ica Aleksand a and Philips Wil ied;
A Ma hema ical Mo phology based
App oach o Vehicle De ec ion in Road
Tunnels; Ghen Uni e si y-TELIN-IPI-
IBBT, Sin -Pie e snieuws aa 41, B-9000
Ghen , Belgium
7. Qian S, Weng G. R.; Resea ch on Objec
De ec ion based on Ma hema ical
Mo phology; In e na ional Con e ence on
In o ma ion Technology and
Managemen Inno a ion (ICITMI 2015)
8. Khu ana Khushboo, Awas hi Ree u;
Techniques o Objec Recogni ion in
Images and Mul i-Objec De ec ion;
In e na ional Jou nal o Ad anced
Resea ch in Compu e Enginee ing and
Technology (IJARCET), Volume 2, Issue
4, Ap il 2013
9. Pei Soo-Chang, Lai Chin-Lun; A
mo phological app oach o a ge
de ec ion on pe spec i e plane; Else ie
Science, Signal P ocessing, 81 (2001)
1975-1984
10. Mu huk ishnan R, M. Radha; Image
de ec ion echnique o Image
segmen a ion; In e na ional Jou nal o
Compu e Science and In o ma ion
Technology (IJCSIT) Volume. 3, No 6,
Dec 2011
11. Hussin R, Rizon Juha i M, Kang Ng Wei,
Ismail R.C.,Kama udin A.; Digi al Image
P ocessing Techniques o Objec
De ec ion F om Complex Backg ound
Image; In e na ional Symposium on
Robo ics and In elligen Senso s 2012
(IRIS 2012)
12. Pa iha Raj a dhan Singh; An Hyb id
Da a Mining App oach o de ec ion and
classi ica ion o Heal h Ca e Da a;
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
M . Akshay Jag ap
367
In e na ional Jou nal o Elec ical,
Elec onics and Compu e Enginee ing
6(1):150-157(2017)
13. Sha ma Fani Bhushan, Sha ma Ajay,
Sha ma Ni mala; Fuzzy logic applica ions
o a ic con ol an op imum and
adap i e con olling applica ion;
In e na ional Jou nal on Eme ging
Technologies,1(1): 41-45 (2010)
14. Jain Ms. Khushbu, Bu se M s. Ka i a,
Mish a M s. Rachna; Con as
Enhancemen o Remo e Sensing Images
Using Op imal Edge De ec ion
Technique; In e na ional Jou nal on
Eme ging Technologies 5(2): 51-
57(2014)
15. Alam M. Zahid, Mish a D . Ra i
Shanka ; A No el wa ele based
h eshold selec ion echnique o Image
Denoising; In e na ional Jou nal on
Eme ging Technologies 3(2): 52-55
(2012)
