Image retrieval using BIM and features from pretrained VGG network for indoor localization

Inhae Ha, Hongjo Kim, Somin Park, Hyoungkwan Kim

Research output: Contribution to journalArticle

9 Citations (Scopus)

Abstract

Various devices that are used indoors require information regarding the user's position and orientation. This information enables the devices to offer the user customized and more relevant information. This study presents a new image-based indoor localization method using building information modeling (BIM) and convolutional neural networks (CNNs). This method constructs a dataset with rendered BIM images and searches the dataset for images most similar to indoor photographs, thereby estimating the indoor position and orientation of the photograph. A pretrained CNN (the VGG network) is used for image feature extraction for the similarity evaluation of two different types of images (BIM rendered and real images). Experiments were performed in real buildings to verify the method, and the matching accuracy is 91.61% for a total of 143 images. The results also confirm that pooling layer 4 in the VGG network is best suited for feature selection.

Original languageEnglish
Pages (from-to)23-31
Number of pages9
JournalBuilding and Environment
Volume140
DOIs
Publication statusPublished - 2018 Aug 1

Fingerprint

Image retrieval
Feature extraction
Neural networks
modeling
neural network
photograph
Experiments
building
experiment
evaluation
method

All Science Journal Classification (ASJC) codes

  • Environmental Engineering
  • Civil and Structural Engineering
  • Geography, Planning and Development
  • Building and Construction

Cite this

@article{8be293e82ed0472a8a467777f42d6f2d,
title = "Image retrieval using BIM and features from pretrained VGG network for indoor localization",
abstract = "Various devices that are used indoors require information regarding the user's position and orientation. This information enables the devices to offer the user customized and more relevant information. This study presents a new image-based indoor localization method using building information modeling (BIM) and convolutional neural networks (CNNs). This method constructs a dataset with rendered BIM images and searches the dataset for images most similar to indoor photographs, thereby estimating the indoor position and orientation of the photograph. A pretrained CNN (the VGG network) is used for image feature extraction for the similarity evaluation of two different types of images (BIM rendered and real images). Experiments were performed in real buildings to verify the method, and the matching accuracy is 91.61{\%} for a total of 143 images. The results also confirm that pooling layer 4 in the VGG network is best suited for feature selection.",
author = "Inhae Ha and Hongjo Kim and Somin Park and Hyoungkwan Kim",
year = "2018",
month = "8",
day = "1",
doi = "10.1016/j.buildenv.2018.05.026",
language = "English",
volume = "140",
pages = "23--31",
journal = "Building and Environment",
issn = "0360-1323",
publisher = "Elsevier BV",

}

Image retrieval using BIM and features from pretrained VGG network for indoor localization. / Ha, Inhae; Kim, Hongjo; Park, Somin; Kim, Hyoungkwan.

In: Building and Environment, Vol. 140, 01.08.2018, p. 23-31.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Image retrieval using BIM and features from pretrained VGG network for indoor localization

AU - Ha, Inhae

AU - Kim, Hongjo

AU - Park, Somin

AU - Kim, Hyoungkwan

PY - 2018/8/1

Y1 - 2018/8/1

N2 - Various devices that are used indoors require information regarding the user's position and orientation. This information enables the devices to offer the user customized and more relevant information. This study presents a new image-based indoor localization method using building information modeling (BIM) and convolutional neural networks (CNNs). This method constructs a dataset with rendered BIM images and searches the dataset for images most similar to indoor photographs, thereby estimating the indoor position and orientation of the photograph. A pretrained CNN (the VGG network) is used for image feature extraction for the similarity evaluation of two different types of images (BIM rendered and real images). Experiments were performed in real buildings to verify the method, and the matching accuracy is 91.61% for a total of 143 images. The results also confirm that pooling layer 4 in the VGG network is best suited for feature selection.

AB - Various devices that are used indoors require information regarding the user's position and orientation. This information enables the devices to offer the user customized and more relevant information. This study presents a new image-based indoor localization method using building information modeling (BIM) and convolutional neural networks (CNNs). This method constructs a dataset with rendered BIM images and searches the dataset for images most similar to indoor photographs, thereby estimating the indoor position and orientation of the photograph. A pretrained CNN (the VGG network) is used for image feature extraction for the similarity evaluation of two different types of images (BIM rendered and real images). Experiments were performed in real buildings to verify the method, and the matching accuracy is 91.61% for a total of 143 images. The results also confirm that pooling layer 4 in the VGG network is best suited for feature selection.

UR - http://www.scopus.com/inward/record.url?scp=85048781544&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85048781544&partnerID=8YFLogxK

U2 - 10.1016/j.buildenv.2018.05.026

DO - 10.1016/j.buildenv.2018.05.026

M3 - Article

AN - SCOPUS:85048781544

VL - 140

SP - 23

EP - 31

JO - Building and Environment

JF - Building and Environment

SN - 0360-1323

ER -