Detekcija ljudi na slikama iz zraka za potrebe misija potrage i spašavanja upotrebom metoda dubokog učenja

Marušić, Željko

prikaz prve stranice dokumenta Detekcija ljudi na slikama iz zraka za potrebe misija potrage i spašavanja upotrebom metoda dubokog učenja

Access restricted to students and staff of home institution

doctoral thesis

Detekcija ljudi na slikama iz zraka za potrebe misija potrage i spašavanja upotrebom metoda dubokog učenja

Split: University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, 2019. urn:nbn:hr:179:298735

Marušić, Željko

University of Split
Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture
Department of Electronics and Computing

Cite this document

APA 6th Edition

Marušić, Ž. (2019). Detekcija ljudi na slikama iz zraka za potrebe misija potrage i spašavanja upotrebom metoda dubokog učenja (Doctoral thesis). Split: University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture. Retrieved from https://urn.nsk.hr/urn:nbn:hr:179:298735

MLA 8th Edition

Marušić, Željko. "Detekcija ljudi na slikama iz zraka za potrebe misija potrage i spašavanja upotrebom metoda dubokog učenja." Doctoral thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, 2019. https://urn.nsk.hr/urn:nbn:hr:179:298735

Chicago 17th Edition

Harvard

Marušić, Ž. (2019). 'Detekcija ljudi na slikama iz zraka za potrebe misija potrage i spašavanja upotrebom metoda dubokog učenja', Doctoral thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, accessed 16 January 2025, https://urn.nsk.hr/urn:nbn:hr:179:298735

Vancouver

Marušić Ž. Detekcija ljudi na slikama iz zraka za potrebe misija potrage i spašavanja upotrebom metoda dubokog učenja [Doctoral thesis]. Split: University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture; 2019 [cited 2025 January 16] Available at: https://urn.nsk.hr/urn:nbn:hr:179:298735

IEEE

Ž. Marušić, "Detekcija ljudi na slikama iz zraka za potrebe misija potrage i spašavanja upotrebom metoda dubokog učenja", Doctoral thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, Split, 2019. Available at: https://urn.nsk.hr/urn:nbn:hr:179:298735

Cite this item: https://urn.nsk.hr/urn:nbn:hr:179:298735

Please login to the repository to save this object to your list.

Metadata

Title	Detekcija ljudi na slikama iz zraka za potrebe misija potrage i spašavanja upotrebom metoda dubokog učenja
Title (english)	Deep Learning Approach in Human Detection on Aerial images for Search and Rescue Missions
Author	Željko Marušić https://orcid.org/0000-0002-1729-6659
Mentor	Sven Gotovac (mentor)
Committee member	Vladan Papić (predsjednik povjerenstva)
Committee member	Sven Gotovac (član povjerenstva)
Committee member	Viktor Sučić (član povjerenstva)
Committee member	Sven Lončarić (član povjerenstva)
Committee member	Tea Marasović (član povjerenstva)
Granter	University of Split Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture (Department of Electronics and Computing) Split
Defense date and country	2019-10-31, Croatia
Scientific / art field, discipline and subdiscipline	TECHNICAL SCIENCES Computing Artificial Intelligence
Universal decimal classification (UDC)	621.3 - Electrical engineering
Abstract	U ovom radu predlaže se novi pristup u detekciji osoba na zračnim slikama prikupljenima s bespilotnih letjelica u misijama potrage i spašavanja u mediteranskim i submediteranskim krajolicima. Detekcija ljudi u slikama visoke prostorne razlučivosti uključuje objekte koji su relativno mali i često teško razaznatljivi u okolišu, stoga je takvo otkrivanje izazovan i zahtjevan zadatak. Predložena metoda počinje smanjivanjem prostora za pretraživanje kroz vizualni algoritam pažnje koji otkriva istaknute ili većinu istaknutih segmenata na slici. Kako bi se smanjio broj nevažnih istaknutih regija, odabrane su one regije koje najvjerojatnije sadrže osobu upotrebom unaprijed obučene i fino podešene konvolucijske neuronske mreže (CNN). Stvorena je posebna baza slika pod nazivom HERIDAL u svrhu treniranja i testiranja predloženog modela. Za akviziciju slika korištene su optičke kamere visoke rezolucije instalirane na bespilotnim letjelicama (UAV). Letovi su se većinom izvodili u relativno niskom letu na neurbanim područjima u prirodi. Za potrebe treniranja prikupljeno je više od 68.750 manjih isječaka odnosno segmenata, a u svrhu testiranje prikupljeno je 500 označenih slika u punoj veličini. Predložena metoda postigla je stopu detekcije od 88,9% i preciznosti od 34,8% čime je pokazana veća učinkovitost od IPSAR sustava kojim se trenutno koriste timovi za potragu i spašavanje u Hrvatskoj. Baza slika HERIDAL je korištena za treniranje i testiranje suvremenog generičkog sustava detekcije Faster R-CNN koji je postigao usporedive, ali ipak nešto lošije rezultate od predložene metode.
Abstract (english)	In this paper, we propose a novel approach to person detection in UAV aerial images for search and rescue tasks in Mediterranean and Sub-Mediterranean landscapes. Person detection in very high spatial resolution images involves target objects that are relatively small and often camouflaged within the environment; thus, such detection is a challenging and demanding task. The proposed method starts by reducing the search space through a visual attention algorithm that detects the salient or most prominent segments in the image. To reduce the number of non-relevant salient regions, we selected those regions most likely to contain a person using pre-trained and fine-tuned convolutional neural networks (CNNs) for detection. We established a special database called HERIDAL to train and test our model. This database was compiled for training purposes, and it contains over 68,750 image patches of wilderness acquired from an aerial perspective as well as approximately 500 labelled full-size real-world images intended for testing purposes. The proposed method achieved a detection rate of 88.9% and a precision of 34.8%, which demonstrates better effectiveness than the system currently used by Croatian Mountain search and rescue (SAR) teams (IPSAR), which is based on mean-shift segmentation. We also used the HERIDAL database to train and test a state-of-the-art region proposal network, Faster R-CNN, which achieved comparable but slightly worse results than those of our proposed method.
Keywords
Keywords (english)
Language	croatian
URN:NBN	urn:nbn:hr:179:298735
Promotion	2021
Study programme	Title: Electrical Engineering and Information Technology Study programme type: university Study level: postgraduate Academic / professional title: doktor/doktorica znanosti, područje tehničkih znanosti, polje elektrotehnika (doktor/doktorica znanosti, područje tehničkih znanosti, polje elektrotehnika)
Type of resource	Text
Extent	146 str.
File origin	Born digital
Access conditions	Access restricted to students and staff of home institution
Terms of use
Repository	Repository of the Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture
Created on	2021-10-07 09:48:06