prikaz prve stranice dokumenta Detekcija vremenskih uvjeta u okolini automobila
  Download
PDF 2.99 MB
master's thesis
Detekcija vremenskih uvjeta u okolini automobila
2019. urn:nbn:hr:200:519652
Triva, Jakob
Josip Juraj Strossmayer University of Osijek
Faculty of Electrical Engineering, Computer Science and Information Technology Osijek
Department of Computer Engineering and Automation
Chair of Automation and Robotics

Cite this document

Triva, J. (2019). Detekcija vremenskih uvjeta u okolini automobila (Master's thesis). Retrieved from https://urn.nsk.hr/urn:nbn:hr:200:519652

Triva, Jakob. "Detekcija vremenskih uvjeta u okolini automobila." Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, 2019. https://urn.nsk.hr/urn:nbn:hr:200:519652

Triva, Jakob. "Detekcija vremenskih uvjeta u okolini automobila." Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, 2019. https://urn.nsk.hr/urn:nbn:hr:200:519652

Triva, J. (2019). 'Detekcija vremenskih uvjeta u okolini automobila', Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, accessed 13 May 2021, https://urn.nsk.hr/urn:nbn:hr:200:519652

Triva J. Detekcija vremenskih uvjeta u okolini automobila [Master's thesis]. Osijek: Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek; 2019 [cited 2021 May 13] Available at: https://urn.nsk.hr/urn:nbn:hr:200:519652

J. Triva, "Detekcija vremenskih uvjeta u okolini automobila", Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, Osijek, 2019. Available at: https://urn.nsk.hr/urn:nbn:hr:200:519652

Islandora Bookmark - Please login to use this feature
Metadata
TitleDetekcija vremenskih uvjeta u okolini automobila
Title (english)Detection of weather conditions around the car
AuthorJakob Triva
MentorRatko Grbić (mentor)
Committee memberRatko Grbić (predsjednik povjerenstva)
Committee memberMario Vranješ (član povjerenstva)
Committee memberMarijan Herceg (član povjerenstva)
GranterJosip Juraj Strossmayer University of Osijek
Faculty of Electrical Engineering, Computer Science and Information Technology Osijek
(Department of Computer Engineering and Automation)
(Chair of Automation and Robotics)
Osijek
Defense date and country2019-09-18, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Artificial Intelligence
Abstract
Problematika kojom se bavi diplomski rad je detekcija vremenskih uvjeta u okolini automobila. U prvom dijelu diplomskog rada opisan je problem detekcije vremenskih uvjeta, kao i neki dosadašnji radovi koji se bave sličnim problemima. Drugim dijelom rada predstavljena je teorijska osnova potrebna za rješavanje praktičnog dijela rada. Rješenje praktičnog dijela rada, kao i rezultati testiranja rješenja opisani su u trećem dijelu diplomskog rada. Za potrebe treniranja klasifikatora izrađeni su skupovi podataka za treniranje i validaciju iz dva izvora: BDD100K baze podataka i izvlačenjem okvira iz prikupljenih video sekvenci. Kreirane su dvije baze podataka: WDC_dataset i DNC_dataset baza podataka, gdje je WDC_dataset korištena pri učenju klasifikatora koji bi trebao raspoznavati vremenske uvjete, a DNC_dataset pri učenju klasifikatora koji bi trebao raspoznavati vrijeme dana, odnosno dan i noć. Predloženo rješenje je realizirano korištenjem konvolucijske neuronske mreže s tri konvolucijska sloja između kojih su korištene ReLu aktivacijske funkcije i slojevi sažimanja po maksimalnoj vrijednosti. Rješenje bazirano na konvolucijskoj neuronskoj mreži je uz korištenje empirijski određenog praga i kružnog međuspremnika za dodatno filtriranje loših predikcija pokazalo zadovoljavajuće rezultate nad testnim video sekvencama. Dobiveno rješenje postiže točnost od 98.3% nad testnim video sekvencama, gdje najbolje rezultate postiže kod detekcije snježnog vremena (100%), a najlošije kod detekcije maglovitog vremena (97.25%).
Abstract (english)
This diploma thesis details the detection of weather conditions in the car environment. The first part of the thesis describes the problem of weather detection, as well as some previous work dealing with same or similar problems. The second part of the paper presents the theoretical basis for dealing with the practical part of the paper. The solution of the practical part of the thesis, as well as the results of testing the solution are described in the third part of the thesis. For training purposes, training and validation data sets were created from two main sources: the BDD100K database and by extracting frames from the collected video sequences. Two databases were created: WDC_dataset database and DNC_dataset database, where WDC_dataset was used for weather conditions classifier training, and DNC_dataset for day-night classifier training. The proposed solution was implemented using convolutional neural network with three convolutional layers, between which ReLu activation functions and max pooling layers were used. The solution based on the convolutional neural network is using a empirically determined threshold and a circular buffer to further filter out false predictions. The proposed solution achieved an accuracy of 98.3% for tested video sequences, where the solution is showing the best results in snowy weather detection (100%) and worst in foggy weather detection (97.25%).
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:200:519652
Study programmeTitle: Graduate University Study Programme in Computer Engineering
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 conditionsOpen access
Terms of useLicence In copyright icon
Created on2019-12-23 11:52:41