prikaz prve stranice dokumenta Nadzor pomorskog prometa u lukama korištenjem algoritama umjetne inteligencije
Access restricted to authenticated users
master's thesis
Nadzor pomorskog prometa u lukama korištenjem algoritama umjetne inteligencije
2021. urn:nbn:hr:179:046157
Matić, Luka
University of Split
Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture
Department of Electronics and Computing

Cite this document

Matić, L. (2021). Nadzor pomorskog prometa u lukama korištenjem algoritama umjetne inteligencije (Master's thesis). Retrieved from https://urn.nsk.hr/urn:nbn:hr:179:046157

Matić, Luka. "Nadzor pomorskog prometa u lukama korištenjem algoritama umjetne inteligencije." Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, 2021. https://urn.nsk.hr/urn:nbn:hr:179:046157

Matić, Luka. "Nadzor pomorskog prometa u lukama korištenjem algoritama umjetne inteligencije." Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, 2021. https://urn.nsk.hr/urn:nbn:hr:179:046157

Matić, L. (2021). 'Nadzor pomorskog prometa u lukama korištenjem algoritama umjetne inteligencije', Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, accessed 27 March 2023, https://urn.nsk.hr/urn:nbn:hr:179:046157

Matić L. Nadzor pomorskog prometa u lukama korištenjem algoritama umjetne inteligencije [Master's thesis]. Split: University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture; 2021 [cited 2023 March 27] Available at: https://urn.nsk.hr/urn:nbn:hr:179:046157

L. Matić, "Nadzor pomorskog prometa u lukama korištenjem algoritama umjetne inteligencije", Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, Split, 2021. Available at: https://urn.nsk.hr/urn:nbn:hr:179:046157

Please login to the repository to save this object to your list.
Metadata
TitleNadzor pomorskog prometa u lukama korištenjem algoritama umjetne inteligencije
Title (english)Ports maritime traffic monitoring using artificial intelligence algorithms
AuthorLuka Matić
MentorSven Gotovac (mentor)
GranterUniversity of Split
Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture
(Department of Electronics and Computing)
Split
Defense date and country2021-07-15, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Abstract
U radu je prikazana realizacija sustava temeljenog na umjetnoj inteligenciji i strojnom učenju za prepoznavanje brodova i brodica u luci Split, sa ciljem nadzora pomorskog prometa. Sustav je baziran na video nadzoru, te je s ciljem zadovoljavajuće pokrivenosti područja luke, odabrana lokacija za smještaj kamere na Koteksovom neboderu. Za algoritam detekcije odabran je YOLOv5, i to njegova inačica YOLOv5x. Izrađena je baza slika za treniranje i videozapisa za testiranje. Baza slika sadržava 324 fotografije od kojih se 256 koristiti za treniranje, a 68 za validaciju modela tijekom treniranja. Slike sadržavaju prikaz svih vremenskih uvjeta i svih doba dana. Treniranjem modela postignuti su sljedeće mjere uspješnosti: mAP_0.5=95.1%, mAP_0.5:0.95=61.91%. Analizirajući ovisnost preciznosti i odziva treniranog modela odabran je optimalni prag pouzdanosti od 60%. Konačno testiranje istreniranog modela provedeno je na brojnim videima, te je utvrđeno kako model detektira sve brodove, te ih ispravno klasificira kao trajekte ili katamarane ili ostale brodove.
Abstract (english)
The paper presents the implementation of a system based on artificial intelligence and machine learning for ship recognition in the port of Split, with the aim of monitoring maritime traffic. The system is based on video surveillance, and to achieve the coverage of the port area, the camera is placed on the Koteks skyscraper. YOLOv5, its version YOLOv5x, was chosen for the detection algorithm. A database of images for training and videos for testing was created. The image database contains 324 photographs of which 256 are used for training and 68 for model validation during training. The images show all weather conditions and all parts of day. The following object detection metrics were achieved by training the model: mAP_0.5 = 95.1%, mAP_0.5: 0.95 = 61.91%. Analysing the dependence of precision and recall of the trained model, the optimal confidence threshold of 60% was selected. The final testing of the trained model was conducted on numerous videos, and it was determined that the model detects all ships, and correctly classifies them as ferries or catamarans or other ships.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:179:046157
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 on2021-11-25 13:24:05