Data-driven scene parsing method for recognizing construction site objects in the whole image

Hongjo Kim, Kinam Kim, Hyoungkwan Kim

Research output: Contribution to journalArticle

14 Citations (Scopus)

Abstract

Although efforts have been made for automated monitoring of construction sites, comprehensive understanding of a whole image remains to be a difficult task. Conventional vision-based monitoring methods have shortcomings in obtaining semantic information regarding an entire image because these methods are not scalable to the number of recognizable objects and training data. Most methods use a parametric model to recognize objects, involving cumbersome parameter tuning. This study presents the data-driven scene parsing method to recognize various objects in a construction site image. For identifying object information of a query image, the monitoring system retrieves the most relevant images to a query image using nearest neighbors and scale invariant feature transform flow matching and transfers relevant image labels to a query image. This study demonstrated the reasonable system performance in construction site images, recording 81.48% of average pixel-wise recognition rate with a small amount of similar images. The scene parsing method would enrich the raw information of a construction site image, thereby facilitating information use for various management applications.

Original languageEnglish
Pages (from-to)271-282
Number of pages12
JournalAutomation in Construction
Volume71
Issue numberPart 2
DOIs
Publication statusPublished - 2016 Nov 1

Fingerprint

Monitoring
Image recording
Information use
Labels
Tuning
Pixels
Semantics

All Science Journal Classification (ASJC) codes

  • Control and Systems Engineering
  • Civil and Structural Engineering
  • Building and Construction

Cite this

@article{ea96b37204e1418bb93d63198d6f98fc,
title = "Data-driven scene parsing method for recognizing construction site objects in the whole image",
abstract = "Although efforts have been made for automated monitoring of construction sites, comprehensive understanding of a whole image remains to be a difficult task. Conventional vision-based monitoring methods have shortcomings in obtaining semantic information regarding an entire image because these methods are not scalable to the number of recognizable objects and training data. Most methods use a parametric model to recognize objects, involving cumbersome parameter tuning. This study presents the data-driven scene parsing method to recognize various objects in a construction site image. For identifying object information of a query image, the monitoring system retrieves the most relevant images to a query image using nearest neighbors and scale invariant feature transform flow matching and transfers relevant image labels to a query image. This study demonstrated the reasonable system performance in construction site images, recording 81.48{\%} of average pixel-wise recognition rate with a small amount of similar images. The scene parsing method would enrich the raw information of a construction site image, thereby facilitating information use for various management applications.",
author = "Hongjo Kim and Kinam Kim and Hyoungkwan Kim",
year = "2016",
month = "11",
day = "1",
doi = "10.1016/j.autcon.2016.08.018",
language = "English",
volume = "71",
pages = "271--282",
journal = "Automation in Construction",
issn = "0926-5805",
publisher = "Elsevier",
number = "Part 2",

}

Data-driven scene parsing method for recognizing construction site objects in the whole image. / Kim, Hongjo; Kim, Kinam; Kim, Hyoungkwan.

In: Automation in Construction, Vol. 71, No. Part 2, 01.11.2016, p. 271-282.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Data-driven scene parsing method for recognizing construction site objects in the whole image

AU - Kim, Hongjo

AU - Kim, Kinam

AU - Kim, Hyoungkwan

PY - 2016/11/1

Y1 - 2016/11/1

N2 - Although efforts have been made for automated monitoring of construction sites, comprehensive understanding of a whole image remains to be a difficult task. Conventional vision-based monitoring methods have shortcomings in obtaining semantic information regarding an entire image because these methods are not scalable to the number of recognizable objects and training data. Most methods use a parametric model to recognize objects, involving cumbersome parameter tuning. This study presents the data-driven scene parsing method to recognize various objects in a construction site image. For identifying object information of a query image, the monitoring system retrieves the most relevant images to a query image using nearest neighbors and scale invariant feature transform flow matching and transfers relevant image labels to a query image. This study demonstrated the reasonable system performance in construction site images, recording 81.48% of average pixel-wise recognition rate with a small amount of similar images. The scene parsing method would enrich the raw information of a construction site image, thereby facilitating information use for various management applications.

AB - Although efforts have been made for automated monitoring of construction sites, comprehensive understanding of a whole image remains to be a difficult task. Conventional vision-based monitoring methods have shortcomings in obtaining semantic information regarding an entire image because these methods are not scalable to the number of recognizable objects and training data. Most methods use a parametric model to recognize objects, involving cumbersome parameter tuning. This study presents the data-driven scene parsing method to recognize various objects in a construction site image. For identifying object information of a query image, the monitoring system retrieves the most relevant images to a query image using nearest neighbors and scale invariant feature transform flow matching and transfers relevant image labels to a query image. This study demonstrated the reasonable system performance in construction site images, recording 81.48% of average pixel-wise recognition rate with a small amount of similar images. The scene parsing method would enrich the raw information of a construction site image, thereby facilitating information use for various management applications.

UR - http://www.scopus.com/inward/record.url?scp=84991062219&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84991062219&partnerID=8YFLogxK

U2 - 10.1016/j.autcon.2016.08.018

DO - 10.1016/j.autcon.2016.08.018

M3 - Article

VL - 71

SP - 271

EP - 282

JO - Automation in Construction

JF - Automation in Construction

SN - 0926-5805

IS - Part 2

ER -