| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

Show document

Title:Vektorizacija rasterskih slik tlorisov zgradb : magistrsko delo
Authors:Oprešnik, Martin (Author)
Potočnik, Božidar (Mentor) More about this mentor... New window
Files:.pdf MAG_Opresnik_Martin_2020.pdf (21,95 MB)
MD5: 2A407004473D80AB1AA239B38B410227
 
Language:Slovenian
Work type:Master's thesis/paper (mb22)
Typology:2.09 - Master's Thesis
Organization:FERI - Faculty of Electrical Engineering and Computer Science
Abstract:V magistrskem delu smo se ukvarjali z vektorizacijo tlorisov stavb. Razvili smo sistem, ki na rasterski sliki tlorisa zazna stene in jih shrani v vektorsko obliko. Omejili smo se na zaznavo sten. Tlorisi lahko poleg označenih sten vsebujejo tudi okna, vrata in pohištvo. Okoli tlorisa je lahko glava načrta, ki vsebuje metapodatke, ki so med drugim naslov, merilo in ime arhitekta. Glavo smo pred vektorizacijo ročno odstranili. Sistem pri večbarvnih načrtih na podlagi barve najprej odstrani oznake, ki ne predstavljajo sten. Za tem zazna tip notacije načrta. Pri načrtih, ki imajo notacijo dveh vzporednih črt s filtriranjem poudarimo črte, sliko binariziramo, izračunamo skeleton slike, zaznamo črte in s pomočjo zaznanih črt zaznamo stene. Pri načrtih z notacijo odebeljene črte pa sliko filtriramo s filtrom mediane, jo binariziramo, zaznamo robove črt in iz zaznanih črt zaznamo stene. Izmed zaznanih sten za tem odstranimo morebiti napačno zaznane stene in izboljšamo natančnost detekcije v vogalih sten. Na koncu zaznane stene shranimo v vektorsko sliko v formatu SVG. Za preizkus naše rešitve smo pripravili podatkovno zbirko 40 tlorisov, od katerih je 20 načrtov za vsak tip notacije. Ročno označene tlorise smo z avtomatsko zaznanimi tlorisi primerjali s pomočjo Jaccardovega indeksa, senzitivnosti, preciznosti in relativne napake seštevka dolžin sten. Na testnih podatkih je naš sistem zaznal stene s povprečnim Jaccardovim indeksom 0,7; senzitivnostjo 0,76; preciznostjo 0,81 in povprečno napako pri seštevku dolžin sten 0,13. Na podlagi rezultatov sklepamo, da je predlagan programski sistem primeren za grobo vektorizacijo, a ni dovolj natančen za popolno avtomatsko uporabo.
Keywords:obdelava slik, tlorisi stavb, vektorizacija
Year of publishing:2020
Place of performance:Maribor
Publisher:[M. Oprešnik]
Number of pages:X, 74 f.
Source:Maribor
UDC:004.932(043.2)
COBISS_ID:44093699 New window
NUK URN:URN:SI:UM:DK:NVD63DLT
Views:119
Downloads:17
Metadata:XML RDF-CHPDL DC-XML DC-RDF
Categories:KTFMB - FERI
:
  
Average score:(0 votes)
Your score:Voting is allowed only for logged in users.
Share:AddThis
AddThis uses cookies that require your consent. Edit consent...

Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Licences

License:CC BY-SA 4.0, Creative Commons Attribution-ShareAlike 4.0 International
Link:http://creativecommons.org/licenses/by-sa/4.0/
Description:This Creative Commons license is very similar to the regular Attribution license, but requires the release of all derivative works under this same license.
Licensing start date:21.10.2020

Secondary language

Language:English
Title:Vectorization of floor plan raster images
Abstract:The subject of this master’s thesis is vectorization of floor plan raster images. We developed a system for detecting walls on raster image of floor plan. We limited ourselves to detect only walls. In addition to walls, floor plans usually also contain windows, doors and furniture. Next to the main part of floor plan there is also a plan header, that contains metadata, such as address, scale and name of architect. Header was manually removed before vectorization. In multi-color plans the system first removes markings that do not represent walls, based on the color of markings. After that it detects the type of notation. In case of notation with two parallel lines, filtering is used to emphasize lines, then image gets binarized and transformed to extract skeleton. Lines are detected on skeleton image and then walls are detected based on these lines. In case of notation with thick lines, image is filtered with median filter and binarized. Walls are detected based on edges found on the image. We remove any possible false detections from detected walls then we improve accuracy in corners. At the end walls are saved in SVG file. In order to test our system we prepared 40 floor plans, 20 with parallel line walls and 20 with thick line walls. We compared automatically vectorized plans with manual vectorization based on Jaccards index, sensitivity, precision and relative lengths sum error. On test dataset our system detected walls with average Jaccard index 0,7; sensitivity 0,76; precision 0,81 and relative lengths sum error 0,13. Based on results we conclude that the system is good enough for rough detection, but is not accurate enough for completely automatic usage.
Keywords:image processing, floor plans, vectorization


Comments

Leave comment

You have to log in to leave a comment.

Comments (0)
0 - 0 / 0
 
There are no comments!

Back
Logos of partners University of Maribor University of Ljubljana University of Primorska University of Nova Gorica