prikaz prve stranice dokumenta Sustav za prepoznavanje bespilotnih letjelica upotrebom konvolucijskih neuronskih mreža
Access restricted to authenticated users
master's thesis
Sustav za prepoznavanje bespilotnih letjelica upotrebom konvolucijskih neuronskih mreža
Split: University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, 2022. urn:nbn:hr:179:683048
Jurić, Toni
University of Split
Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture
Department of Electronics and Computing

Institutional repository: Repository of the Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture

Cite this document

Jurić, T. (2022). Sustav za prepoznavanje bespilotnih letjelica upotrebom konvolucijskih neuronskih mreža (Master's thesis). Split: University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture. Retrieved from https://urn.nsk.hr/urn:nbn:hr:179:683048

Jurić, Toni. "Sustav za prepoznavanje bespilotnih letjelica upotrebom konvolucijskih neuronskih mreža." Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, 2022. https://urn.nsk.hr/urn:nbn:hr:179:683048

Jurić, Toni. "Sustav za prepoznavanje bespilotnih letjelica upotrebom konvolucijskih neuronskih mreža." Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, 2022. https://urn.nsk.hr/urn:nbn:hr:179:683048

Jurić, T. (2022). 'Sustav za prepoznavanje bespilotnih letjelica upotrebom konvolucijskih neuronskih mreža', Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, accessed 29 June 2024, https://urn.nsk.hr/urn:nbn:hr:179:683048

Jurić T. Sustav za prepoznavanje bespilotnih letjelica upotrebom konvolucijskih neuronskih mreža [Master's thesis]. Split: University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture; 2022 [cited 2024 June 29] Available at: https://urn.nsk.hr/urn:nbn:hr:179:683048

T. Jurić, "Sustav za prepoznavanje bespilotnih letjelica upotrebom konvolucijskih neuronskih mreža", Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, Split, 2022. Available at: https://urn.nsk.hr/urn:nbn:hr:179:683048

Metadata
TitleSustav za prepoznavanje bespilotnih letjelica upotrebom konvolucijskih neuronskih mreža
Title (english)Unmanned aerial vehicle recognition system using convolutional neural networks
AuthorToni Jurić
MentorIvo Stančić https://orcid.org/0000-0001-9607-1791 (mentor)
GranterUniversity of Split
Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture
(Department of Electronics and Computing)
Split
Defense date and country2022-07-13, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Abstract
Ovaj rad prikazuje rješenje problema detekcije dronova korištenjem neuronskih mreža primijenjenih na slikovnim signalima. U tu svrhu obrađen je način rada konvolucijskih neuronskih mreža koje se koriste za sve vrste zadataka vezane uz analizu slike kao što su klasifikacija, detekcija objekata i segmentacija. Prvo su predstavljene neke od najpoznatijih klasifikacijskih neuronskih mreža: LeNet, AlexNet, VGG, ResNet i MobileNet, s ciljem prikazivanja razvoja arhitektura i metoda učenja mreža koje se i danas koriste. Zatim su prikazane najpoznatije neuronske mreže za detekciju objekata: R-CNN, SSD i YOLO koje korištenjem jednih od istaknutih klasifikacijskih neuronskih mreža omogućavaju i lokalizaciju objekata na slici. Navedene neuronske mreže za detekciju primijenjene su na vlastitom skupu podataka za detekciju 5 klasa objekata: dronova, ptica, zrakoplova, helikoptera i insekata. Dobiveni rezultati se dodatno poboljšavaju korištenjem sintetičkih podataka generiranih u Blenderu. Kako bi se ispitala mogućnost izvršavanja odabranih neuronskih mreža u realnom vremenu, testirane su njihove performanse na različitim uređajima: Raspberry Pi 4B, Jetson TX2 i stolnom računalu. Konačno, s ciljem dodatnog poboljšanja rezultata klasifikacije koristi se detekcija u dva koraka.
Abstract (english)
This paper illustrates one possible solution for the problem of drone detection by applying neural networks on images. For this purpose, convolutional neural networks are discussed which are used for all types of tasks related to image analysis such as classification, object detection and segmentation. Firstly, some of the most popular classification neural networks are presented including: LeNet, AlexNet, VGG, ResNet and MobileNet, with the aim of showing the development of network architectures and training methods that are used today. Then the most common neural networks for object detection are presented including: R-CNN, SSD and YOLO which, by using one of the already mentioned classification networks, enable the localization of objects in images. These neural networks for object detection are then trained on a collected dataset for detecting 5 classes of objects: drones, birds, airplanes, helicopters and insects. The obtained results are further improved by using synthetic data generated in Blender. Furthermore, in order to examine the possibility of executing the selected neural networks in real time, their performance was tested on different devices including: Raspberry Pi 4B, Jetson TX2 and desktop computer. Finally, twostage detection is used to further improve the classification results.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:179:683048
Study programmeTitle: Computing
Study programme type: university
Study level: graduate
Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
Type of resourceText
File originBorn digital
Access conditionsAccess restricted to authenticated users
Terms of useLicence In copyright icon
Created on2023-05-18 12:52:03