ACTIVE RESEARCHER
SCIENTIFIC JOURNAL
ac i e esea che .com Noyab , 2025 olume 2 issue 11
24
DOI: h ps://doi.o g/10.5281/zenodo.17711157
THE ROLE OF COMPUTER VISION IN ROBOTIC SYSTEMS
Yuldashe Nodi bek Olimo ich
FGTU, Mas e ’s S uden
meda o @mail. u
Abdukadi o Abdu akhi Gapi o ich
FGTU, Associa e P o esso , PhD
a_ ahy [email protected]
ABSTRACT
The esea ch in es iga es how compu e ision unc ions in obo ic sys ems o
p o ide au onomous en i onmen al pe cep ion and decision capabili ies and
in e ac ion abili ies. The esea ch in es iga es he undamen al me hods and sys em
designs and app oaches which enable con empo a y obo s o achie e isual
pe cep ion. The pape ocuses on eal-wo ld applica ions o au onomous na iga ion
and objec ecogni ion and manipula ion asks.
Keywo ds: Compu e ision, obo ics, au onomous na iga ion, deep lea ning,
image p ocessing.
In oduc ion
Robo ics de elopmen has ecei ed subs an ial impac om compu e ision
echnology imp o emen s. Robo s use compu e ision o unde s and hei
en i onmen h ough came a and senso da a analysis which enables hem o in e p e
isual in o ma ion. The abili y o in e p e isual da a enables obo s o pe o m
au onomous na iga ion and objec manipula ion and make in elligen decisions. Visual
pe cep ion in eg a ion enables obo s o execu e complex asks in unp edic able
se ings which minimizes he equi emen o human o e sigh .
ACTIVE RESEARCHER
SCIENTIFIC JOURNAL
ac i e esea che .com Noyab , 2025 olume 2 issue 11
25
Me hods
Robo ics applica ions o compu e ision equi e ha dwa e and so wa e elemen s
ha ope a e in eal ime o achie e pe cep ion. The p ocess o obo ics pe cep ion
in ol es ou essen ial s ages which s a wi h image acquisi ion ollowed by ea u e
ex ac ion and hen objec de ec ion and seman ic unde s anding. Mode n app oaches
use deep lea ning models wi h con olu ional neu al ne wo ks (CNNs) o pe o m
obus ea u e lea ning and classi ica ion ope a ions. The combina ion o senso usion
echnology be ween came as and LiDAR senso s and IMU senso s p oduces be e
accu acy and dep h pe cep ion esul s.
Table 1 — Compa ison o Compu e Vision Algo i hms
Algo i hm
Func ion
Ad an ages
Limi a ions
CNN
Fea u e ex ac ion
and classi ica ion
High accu acy
Requi es la ge
da ase s
YOLO
Real- ime objec
de ec ion
Fas p ocessing
Lowe p ecision in
small objec s
SIFT
Fea u e ma ching
In a ian o scale
and o a ion
Compu a ionally
expensi e
Op ical Flow
Mo ion es ima ion
Wo ks in eal- ime
Sensi i e o
ligh ing changes
Resul s
Robo ics achie es be e au onomy and pe o mance h ough expe imen al
compu e ision implemen a ions. Robo s ha use ision-based na iga ion sys ems can
pe o m obs acle de ec ion and objec ecogni ion and en i onmen al adap a ion.
Vision-guided obo ic a ms in manu ac u ing ope a ions achie e be e assembly line
p ecision and compu e ision enables sa e adap i e d i ing o au onomous ehicles.
ACTIVE RESEARCHER
SCIENTIFIC JOURNAL
ac i e esea che .com Noyab , 2025 olume 2 issue 11
26
Table 2 — Applica ions o Compu e Vision in Robo ics
Applica ion A ea
Func ionali y
Example
Au onomous
Na iga ion
En i onmen mapping
and pa h planning
Sel -d i ing ehicles
Indus ial Robo ics
Pa inspec ion and
assembly
Vision-guided obo ic
a ms
Medical Robo ics
Su gical p ecision and
diagnosis
Robo -assis ed su ge y
Ag icul u al Robo ics
C op moni o ing and
ha es ing
Vision-based d ones
Discussion
Wi h he de elopmen o compu e ision echnology in obo ic sys ems,
au oma ion and in elligence ha e also made g ea p og ess. The e a e di icul ies ha
es ablished a exceed any hing encoun e ed by ANNs, especially wi h espec o eal-
ime p ocessing in unce ain and dynamic en i onmen s. I is expec ed ha u u e
b eak h oughs may be achie ed by in eg a ing ision-based AI wi h edge compu ing
and neu omo phic ha dwa e o as e and mo e ene gy-e icien p ocessing.
Conclusion
Compu e ision is a co e enabling echnology in he obo ics, allowing machines
o see, in e p e and ac wi hou human in e en ion. I s ole is g owing in indus ies
anging om au onomous ehicles o heal hca e and manu ac u ing. Cu en esea ch
ocuses on inc easing obus ness, accu acy and in e p e abili y and os e s he
de elopmen o uly in elligen obo ic sys ems.
ACTIVE RESEARCHER
SCIENTIFIC JOURNAL
ac i e esea che .com Noyab , 2025 olume 2 issue 11
27
REFERENCES
[1] S. Th un, W. Bu ga d, and D. Fox, *P obabilis ic Robo ics*, MIT P ess, 2005.
[2] K. Simonyan and A. Zisse man, “Ve y Deep Con olu ional Ne wo ks o La ge-
Scale Image Recogni ion,” a Xi :1409.1556, 2014.
[3] Redmon, J. e al., “You Only Look Once: Uni ied, Real-Time Objec De ec ion,”
CVPR, 2016.
[4] H. Hi schmulle , “S e eo P ocessing by Semiglobal Ma ching and Mu ual
In o ma ion,” IEEE TPAMI, 2008.
