izpis_h1_title_alt

Automatic building footprint extraction from UAV images using neural networks
ID Kokeza, Zoran (Author), ID Vujasinović, Miroslav (Author), ID Govedarica, Miro (Author), ID Milojević, Brankica (Author), ID Jakovljević, Gordana (Author)

.pdfPDF - Presentation file, Download (2,89 MB)
MD5: 285CB121BBEAF8C564C04D96757E0FC9

Abstract
Up-to-date cadastral maps are crucial for urban planning. Creating those maps with the classical geodetic methods is expensive and time-consuming. Emerge of Unmanned Aerial Vehicles (UAV) made a possibility for quick acquisition of data with much more details than it was possible before. The topic of the research refers to the challenges of automatic extraction of building footprints on high-resolution orthophotos. The objectives of this study were as follows: (1) to test the possibility of using different publicly available datasets (Tanzania, AIRS and Inria) for neural network training and then test the generalisation capability of the model on the Area Of Interest (AOI); (2) to evaluate the effect of the normalised digital surface model (nDSM) on the results of neural network training and implementation. Evaluation of the results shown that the models trained on the Tanzania (IoU 36.4%), AIRS (IoU 64.4%) and Inria (IoU 7.4%) datasets doesn't satisfy the requested accuracy to update cadastral maps in study area. Much better results are achieved in the second part of the study, where the training of the neural network was done on tiles (256x256) of the orthophoto of AOI created from data acquired using UAV. A combination of RGB orthophoto with nDSM resulted in a 2% increase of IoU, achieving the final IoU of over 90%.

Language:English
Keywords:neural network, deep learning, classification, Structure from Motion, unmanned aerial vehicles, building footprint extraction
Work type:Scientific work
Typology:1.01 - Original Scientific Article
Organization:FGG - Faculty of Civil and Geodetic Engineering
Publication status:Published
Publication version:Version of Record
Article acceptance date:18.02.2020
Publication date:26.10.2020
Year:2020
Number of pages:Str. 545-561
Numbering:Letn. 64, št. 4
PID:20.500.12556/RUL-132797 This link opens in a new window
UDC:004.032.26:528.7/.8
ISSN on article:0351-0271
DOI:10.15292/geodetski-vestnik.2020.04.545-561 This link opens in a new window
COBISS.SI-ID:45179651 This link opens in a new window
Publication date in RUL:03.11.2021
Views:782
Downloads:116
Metadata:XML RDF-CHPDL DC-XML DC-RDF
:
Copy citation
Share:Bookmark and Share

Record is a part of a journal

Title:Geodetski vestnik. glasilo Zveze geodetov Slovenije
Shortened title:Geod. vestn.
Publisher:Zveza geodetov Slovenije
ISSN:0351-0271
COBISS.SI-ID:5091842 This link opens in a new window

Licences

License:CC BY 4.0, Creative Commons Attribution 4.0 International
Link:http://creativecommons.org/licenses/by/4.0/
Description:This is the standard Creative Commons license that gives others maximum freedom to do what they want with the work as long as they credit the author.
Licensing start date:26.10.2020

Secondary language

Language:Slovenian
Title:Samodejno zajemanje odtisov stavb iz UAV podob z uporabo nevronskih mrež
Abstract:
Ažurni katastrski načrti so ključnega pomena za urbanistično načrtovanje, njihovo posodabljanje pa je drago in zahteva veliko časa. Razvoj daljinsko vodenih letalnikov (angl. unmanned aerial vehicles % UAV) je omogočil hiter zajem podatkov z veliko višjo stopnjo podrobnosti od klasične geodetske izmere. V raziskavi se ukvarjamo s samodejnim zajemom odtisov stavb iz ortofotov visoke ločljivosti. Cilja naše študije sta bila: (1) preskusiti možnosti uporabe različnih javno dostopnih podatkovnih nizov (Tanzania, AIRS in Inria) za učenje nevronskih mrež ter nato preskusiti zmožnosti generalizacije modela območja obravnave; (2) oceniti vpliv normaliziranega digitalnega modela površja na rezultate učenja in uporabo nevronskih mrež. Rezultati so pokazali, da modeli, ki smo jih učili na omenjenih podatkovnih nizih, niso zadovoljivi, saj je prekrivanje identificiranih odtisov stavb z referenčnimi podatki znašalo za podatkovni niz Tanzania 36,4%, za AIRS je bila vrednost 64,4% za Inria pa le 7,4%. Boljše rezultate smo dosegli v drugem delu raziskave, kjer je bilo učenje nevronskih mrež izvedeno na delih (256 x 256 pikslov) ortofota, pridobljenega na podlagi podatkov, zajetih z UAV. Pri kombiniranju ortofota z normaliziranim digitalnim modelom površja se je še povečal delež prostorskega ujemanja z referenčnimi podatki (IoU) in znašalo 90 %.

Keywords:nevronske mreže, globoko učenje, klasifikacija, grajenje strukture iz gibanja, daljinsko vodeni letalniki, zajemanje odtisa stavbe

Similar documents

Similar works from RUL:
Similar works from other Slovenian collections:

Back