diplomsko delo
Aleš Kert (Author), Matej Kristan (Mentor)

Abstract

V diplomskem delu predstavimo problem detekcije poškodb na steklu. Predlagamo novo metodo, ki se namesto zanašanja na slike klasičnih kamer, zanaša na podatke o polarizaciji, pridobljenih s polarizacijsko kamero. Tekom razvoja metode je bila zajeta prva javno dostopna podatkovna zbirka, ki vsebuje polarizacijske slike poškodb na vetrobranskih steklih. Na zbirki so nato ročno označene poškodbe. Detekcijo poškodb stekla opišemo kot problem semantične segmentacije, kjer za vsak piksel določimo, če predstavlja poškodbo ali ne. Reševanja problema se lotimo z uporabo konvolucijskih nevronskih mrež. Pri izbiri vhodnih podatkov mreže eksperimentiramo z različnimi obdelavami polarizacijskih slik, z namenom odkritja optimalne obdelave. Najboljše rezultate nam nudi model, ki za vhod prejme nespremenjene polarizacijske slike, zložene v štiri kanale. Ta model dosega natančnost 0.923, priklic 0.861 in F-mero 0.885.

Keywords

polarizacija;polarizacijska kamera;konvolucijska nevronska mreža;semantična segmentacija;računalništvo in informatika;računalništvo in matematika;interdisciplinarni študij;univerzitetni študij;diplomske naloge;

Data

Language: Slovenian
Year of publishing:
Typology: 2.11 - Undergraduate Thesis
Organization: UL FRI - Faculty of Computer and Information Science
Publisher: [A. Kert]
UDC: 004(043.2)
COBISS: 76733699 Link will open in a new window
Views: 719
Downloads: 55
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Detection of glass damage using a polarization camera
Secondary abstract: This thesis addresses the problem of detecting glass damage. We propose a new method that uses polarization data captured with a polarization camera, instead of relying on pictures taken with a classic camera. During the development of this method, we created the first publicly available dataset containing polarization pictures of windshield damage. The damage was annotated by hand. We pose the detection of glass damage as a semantic segmentation problem, where each pixel is classified as either healthy or damaged. We try to solve this problem using convolutional neural networks. Furthermore, we evaluate different ways of processing polarization pictures to determine the optimal processing method. The best results are offered by a model, that uses unchanged polarization images, arranged into a four channel image, which achieves a precision of 0.923, recall of 0.861 and F-score of 0.885.
Secondary keywords: polarization;polarization camera;convolutional neural network;semantic segmentation;computer science;computer and information science;computer science and mathematics;interdisciplinary studies;diploma;Računalništvo;Univerzitetna in visokošolska dela;
Type (COBISS): Bachelor thesis/paper
Study programme: 1000407
Embargo end date (OpenAIRE): 1970-01-01
Thesis comment: Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages: 73 str.
ID: 13345768