Semantička segmentacija RGB-D slike unutrašnjih prostora

Šimić, Kristijan

master's thesis

Semantička segmentacija RGB-D slike unutrašnjih prostora

Osijek: Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, 2015. urn:nbn:hr:200:551948

Šimić, Kristijan

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

APA 6th Edition

Šimić, K. (2015). Semantička segmentacija RGB-D slike unutrašnjih prostora (Master's thesis). Osijek: Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek. Retrieved from https://urn.nsk.hr/urn:nbn:hr:200:551948

MLA 8th Edition

Šimić, Kristijan. "Semantička segmentacija RGB-D slike unutrašnjih prostora." Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, 2015. https://urn.nsk.hr/urn:nbn:hr:200:551948

Chicago 17th Edition

Harvard

Šimić, K. (2015). 'Semantička segmentacija RGB-D slike unutrašnjih prostora', Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, accessed 26 April 2024, https://urn.nsk.hr/urn:nbn:hr:200:551948

Vancouver

Šimić K. Semantička segmentacija RGB-D slike unutrašnjih prostora [Master's thesis]. Osijek: Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek; 2015 [cited 2024 April 26] Available at: https://urn.nsk.hr/urn:nbn:hr:200:551948

IEEE

K. Šimić, "Semantička segmentacija RGB-D slike unutrašnjih prostora", Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, Osijek, 2015. Available at: https://urn.nsk.hr/urn:nbn:hr:200:551948

Cite this item: https://urn.nsk.hr/urn:nbn:hr:200:551948

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

Metadata

Title	Semantička segmentacija RGB-D slike unutrašnjih prostora
Title (english)	Semantic Segmentation of RGB-D Images of Indoor Environments
Author	Kristijan Šimić
Mentor	Robert Cupec (mentor)
Committee member	Robert Cupec (predsjednik povjerenstva)
Committee member	Emmanuel Karlo Nyarko (član povjerenstva)
Committee member	Ratko Grbić (član povjerenstva)
Granter	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) Osijek
Defense date and country	2015-12-17, Croatia
Scientific / art field, discipline and subdiscipline	TECHNICAL SCIENCES Computing Process Computing
Abstract	U ovom radu je analiziran jedan pristup rješavanju problema semantičke segmentacije dubinskih slika unutrašnjih prostora. Proučeni su algoritmi korišteni u predloženoj metodi, kao i njihove programske implementacije. Rješavanjem problema semantičke segmentacije slike znatno bi se poboljšala sposobnost robota za samostalan rad u nepoznatim okruženjima. Problem se svodi na svrstavanje dijelova slike u neke unaprijed određene kategorije. U ovom slučaju te kategorije su pod, strukture, namještaj i rekviziti. Predložena metoda kategorizira dijelove slike koristeći uvjetna slučajna polja. Slika se najprije segmentira na superpiksele, koji služe kao čvorovi grafa uvjetnog slučajnog polja. Potencijali uvjetnog slučajnog polja određeni su vizualnim i geometrijskim značajkama superpiksela. Struktura grafa je minimalno razapinjuće stablo, pri čemu je težina brida određena na temelju udaljenosti centroida dvaju superpiksela u trodimenzionalnom prostoru. U okviru ovog rada proučena je jedna metoda zaključivanja i dekodiranja uvjetnih slučajnih polja koja omogućuje pronalaženje optimalne segmentacije. Učinkovitost predložene metode ispitana je eksperimentima na ispitnom skupu dubinskih slika unutrašnjih prostora.
Abstract (english)	This thesis considers an approach for semantic segmentation of RGB-D images of indoor environments. The algorithms used in this approach are described and their implementation in the provided source code is explained. Solving the problem of semantic RGB-D image segmentation would significantly improve the capacity of robots for autonomous problem solving in unknown environments. The considered problem is to classify parts of an image into a set of predefined categories. In this case, the categories are ground, structure, furniture and props. The discussed method does this using the conditional random fields. The image is first segmented into superpixels, which serve as nodes in a CRF graph, where the CRF potentials are defined by visual and geometric features of the superpixels that the image is divided into. The structure of the CRF graph is a minimum spanning tree, with the weight of the graph edges being defined by the distance between the superpixel centroids in three-dimensional space. One method for CRF inference and decoding, which provides an optimal segmentation, is examined. The efficacy of the considered approach is tested on a set of RGB-D images of indoor environments.
Keywords
Keywords (english)
Language	croatian
URN:NBN	urn:nbn:hr:200:551948
Study programme	Title: 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 resource	Text
File origin	Born digital
Access conditions	Open access
Terms of use
Repository	Faculty of Electrical Engineering, Computer Science and Information Technology Osijek
Created on	2016-08-01 11:31:26