prikaz prve stranice dokumenta Detekcija i praćenje osoba u video zapisima temeljena na HOG značajkama
Access restricted to students and staff of home institution
undergraduate thesis
Detekcija i praćenje osoba u video zapisima temeljena na HOG značajkama
Rijeka: University of Rijeka, Faculty of Engineering, 2020. urn:nbn:hr:190:271469
Gardijan, Andrej
University of Rijeka
Faculty of Engineering
Department of Computer Engineering
Section of Intelligent Computing Systems

Institutional repository: Repository of the University of Rijeka, Faculty of Engineering

Cite this document

Gardijan, A. (2020). Detekcija i praćenje osoba u video zapisima temeljena na HOG značajkama (Undergraduate thesis). Rijeka: University of Rijeka, Faculty of Engineering. Retrieved from https://urn.nsk.hr/urn:nbn:hr:190:271469

Gardijan, Andrej. "Detekcija i praćenje osoba u video zapisima temeljena na HOG značajkama." Undergraduate thesis, University of Rijeka, Faculty of Engineering, 2020. https://urn.nsk.hr/urn:nbn:hr:190:271469

Gardijan, Andrej. "Detekcija i praćenje osoba u video zapisima temeljena na HOG značajkama." Undergraduate thesis, University of Rijeka, Faculty of Engineering, 2020. https://urn.nsk.hr/urn:nbn:hr:190:271469

Gardijan, A. (2020). 'Detekcija i praćenje osoba u video zapisima temeljena na HOG značajkama', Undergraduate thesis, University of Rijeka, Faculty of Engineering, accessed 02 May 2024, https://urn.nsk.hr/urn:nbn:hr:190:271469

Gardijan A. Detekcija i praćenje osoba u video zapisima temeljena na HOG značajkama [Undergraduate thesis]. Rijeka: University of Rijeka, Faculty of Engineering; 2020 [cited 2024 May 02] Available at: https://urn.nsk.hr/urn:nbn:hr:190:271469

A. Gardijan, "Detekcija i praćenje osoba u video zapisima temeljena na HOG značajkama", Undergraduate thesis, University of Rijeka, Faculty of Engineering, Rijeka, 2020. Available at: https://urn.nsk.hr/urn:nbn:hr:190:271469

Metadata
TitleDetekcija i praćenje osoba u video zapisima temeljena na HOG značajkama
Title (english)Detection and Person Tracking in videos based on HOG features
AuthorAndrej Gardijan
MentorIvo Ipšić (mentor)
Committee memberKristijan Lenac (predsjednik povjerenstva)
Committee memberIvo Ipšić (član povjerenstva)
GranterUniversity of Rijeka
Faculty of Engineering
(Department of Computer Engineering)
(Section of Intelligent Computing Systems)
Rijeka
Defense date and country2020-09-16, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Abstract
U ovom završnom radu prikazan je princip rada HOG metode uz programski kod koji prikazuje primjenu iste u detekciji i praćenju osoba u video zapisima. Prikazana je pozadina iza HOG opisnika, značenje njega samog kao vektora histograma orijentiranih gradijenata. Iz podataka znanstvenog rada Dalal i Triggs vidljivi su optimalni parametri za detekciju raznih objekata pa tako i osoba. Sam proces detekcije razbijen je na nekoliko koraka od kojih svaki igra bitnu ulogu i doprinosi kvalitetnijoj detekciji osoba. Takoder je prikazana i stopa promašaja detekcije u ovisnosti o veličini ćelije i veličine blokova uz prisutnost preklapanja blokova. Kako bi se poboljšala kvaliteta detekcija, a tako i praćenja te smanjila redundancija korišteno je non-max suzbijanje koje osigurava jednostruku detekciju objekta, u ovom slučaju osobe. Tijekom praćenja korištena je metoda praćenja centroida, koja na temelju euklidske udaljenosti vrši jednostavno praćenje detektiranih četverokuta. Primjena svega opisanog vidljiva je na setu podataka CAVIAR.
Abstract (english)
This final thesis shows the principles behind the HOG method along with the source code which shows the application of the mentioned method in people detection and tracking in video recordings. The background workings of the HOG descriptor are shown, along with its meaning as a vector of histogram of oriented gradients. The optimal parameters for detection of objects as well as people are visible from Dalal’s and Triggs’s thesis. The detection process is broken down into smaller steps of which each one plays an important role and contributes to its people detection quality. Among other things the miss rate is shown, its dependency on cell size as well as block size and its overlap. To improve the detection quality and lower redundancy non-max suppression was used which ensures a single detection of an object, in this case a person. The tracking method used was centroid tracking which simply based on euclidean distance tracks the detected quadrangles. The application of all that is mentioned is visible on the CAVIAR data set.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:190:271469
Study programmeTitle: Computer Engineering
Study programme type: university
Study level: undergraduate
Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
Type of resourceText
File originBorn digital
Access conditionsAccess restricted to students and staff of home institution
Terms of useLicence In copyright icon
Created on2021-09-30 09:06:06