Transferability of object-based image analysis approaches for landslide detection in the Himalaya Mountains of northern Pakistan

Alam Sher Bacha*, H.M.A. Van Der Werff, Muhammad Shafique, Hawas Khan

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

Abstract

A landslide inventory is indispensable for determination of landslide susceptibility, hazard, risk assessment and disaster mitigation strategies. These inventories were traditionally developed using manual digitization of remote sensing images and aerial photographs, and pixel-based image classification. Recently, Object-Based Image Analysis (OBIA) supersedes visual interpretation and pixel-based methods. OBIA utilizes spectral, textural, contextual, morphological and topographical information in remote sensing images. However, OBIA-based landslide detection methods are often designed for specific areas and remote sensing dataset. The aim of this study is to evaluate the transferability of three published OBIA landslide detection methods for semi-automated landslide detection in the Himalaya mountainous region of northern Pakistan. A SPOT-6 multi-spectral image with Advanced Land Observing Satellite (ALOS) Phased Array type L-band Synthetic Aperture Radar (PALSAR) Digital Elevation Model (DEM) derivatives, i.e. slope, aspect, hill-shade, relief, elevation and stream network are used for landslide detection using eCognition developer software. The three published methods scale parameters for image segmentation and parameter thresholds are evaluated first. It is observed that the aforementioned methods are not directly applicable to our study area and remote sensing datasets. Therefore, an alternate (proposed) method is developed for semi-automated landslide detection. Accuracy assessment of the selected methods and proposed method is assessed by Precision, Recall and F1 measures. Using the proposed method, a total of 357 landslides are detected with 91.46% Precision, 93.31% Recall and 92.38% F1 measure accuracy.

Original languageEnglish
Pages (from-to)3390-3410
Number of pages21
JournalInternational journal of remote sensing
Volume41
Issue number9
Early online date2 Jan 2020
DOIs
Publication statusE-pub ahead of print/First online - 2 Jan 2020

Fingerprint

image analysis
landslide
mountain
remote sensing
detection method
pixel
PALSAR
digitization
accuracy assessment
multispectral image
detection
image classification
SPOT
method
aerial photograph
segmentation
digital elevation model
disaster
risk assessment
relief

Keywords

  • ITC-ISI-JOURNAL-ARTICLE
  • UT-Hybrid-D

Cite this

@article{b9a12aa84d644cf89d1cd89335669492,
title = "Transferability of object-based image analysis approaches for landslide detection in the Himalaya Mountains of northern Pakistan",
abstract = "A landslide inventory is indispensable for determination of landslide susceptibility, hazard, risk assessment and disaster mitigation strategies. These inventories were traditionally developed using manual digitization of remote sensing images and aerial photographs, and pixel-based image classification. Recently, Object-Based Image Analysis (OBIA) supersedes visual interpretation and pixel-based methods. OBIA utilizes spectral, textural, contextual, morphological and topographical information in remote sensing images. However, OBIA-based landslide detection methods are often designed for specific areas and remote sensing dataset. The aim of this study is to evaluate the transferability of three published OBIA landslide detection methods for semi-automated landslide detection in the Himalaya mountainous region of northern Pakistan. A SPOT-6 multi-spectral image with Advanced Land Observing Satellite (ALOS) Phased Array type L-band Synthetic Aperture Radar (PALSAR) Digital Elevation Model (DEM) derivatives, i.e. slope, aspect, hill-shade, relief, elevation and stream network are used for landslide detection using eCognition developer software. The three published methods scale parameters for image segmentation and parameter thresholds are evaluated first. It is observed that the aforementioned methods are not directly applicable to our study area and remote sensing datasets. Therefore, an alternate (proposed) method is developed for semi-automated landslide detection. Accuracy assessment of the selected methods and proposed method is assessed by Precision, Recall and F1 measures. Using the proposed method, a total of 357 landslides are detected with 91.46{\%} Precision, 93.31{\%} Recall and 92.38{\%} F1 measure accuracy.",
keywords = "ITC-ISI-JOURNAL-ARTICLE, UT-Hybrid-D",
author = "Bacha, {Alam Sher} and {Van Der Werff}, H.M.A. and Muhammad Shafique and Hawas Khan",
year = "2020",
month = "1",
day = "2",
doi = "10.1080/01431161.2019.1701725",
language = "English",
volume = "41",
pages = "3390--3410",
journal = "International journal of remote sensing",
issn = "0143-1161",
publisher = "Taylor & Francis",
number = "9",

}

Transferability of object-based image analysis approaches for landslide detection in the Himalaya Mountains of northern Pakistan. / Bacha, Alam Sher; Van Der Werff, H.M.A.; Shafique, Muhammad; Khan, Hawas.

In: International journal of remote sensing, Vol. 41, No. 9, 02.05.2020, p. 3390-3410.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - Transferability of object-based image analysis approaches for landslide detection in the Himalaya Mountains of northern Pakistan

AU - Bacha, Alam Sher

AU - Van Der Werff, H.M.A.

AU - Shafique, Muhammad

AU - Khan, Hawas

PY - 2020/1/2

Y1 - 2020/1/2

N2 - A landslide inventory is indispensable for determination of landslide susceptibility, hazard, risk assessment and disaster mitigation strategies. These inventories were traditionally developed using manual digitization of remote sensing images and aerial photographs, and pixel-based image classification. Recently, Object-Based Image Analysis (OBIA) supersedes visual interpretation and pixel-based methods. OBIA utilizes spectral, textural, contextual, morphological and topographical information in remote sensing images. However, OBIA-based landslide detection methods are often designed for specific areas and remote sensing dataset. The aim of this study is to evaluate the transferability of three published OBIA landslide detection methods for semi-automated landslide detection in the Himalaya mountainous region of northern Pakistan. A SPOT-6 multi-spectral image with Advanced Land Observing Satellite (ALOS) Phased Array type L-band Synthetic Aperture Radar (PALSAR) Digital Elevation Model (DEM) derivatives, i.e. slope, aspect, hill-shade, relief, elevation and stream network are used for landslide detection using eCognition developer software. The three published methods scale parameters for image segmentation and parameter thresholds are evaluated first. It is observed that the aforementioned methods are not directly applicable to our study area and remote sensing datasets. Therefore, an alternate (proposed) method is developed for semi-automated landslide detection. Accuracy assessment of the selected methods and proposed method is assessed by Precision, Recall and F1 measures. Using the proposed method, a total of 357 landslides are detected with 91.46% Precision, 93.31% Recall and 92.38% F1 measure accuracy.

AB - A landslide inventory is indispensable for determination of landslide susceptibility, hazard, risk assessment and disaster mitigation strategies. These inventories were traditionally developed using manual digitization of remote sensing images and aerial photographs, and pixel-based image classification. Recently, Object-Based Image Analysis (OBIA) supersedes visual interpretation and pixel-based methods. OBIA utilizes spectral, textural, contextual, morphological and topographical information in remote sensing images. However, OBIA-based landslide detection methods are often designed for specific areas and remote sensing dataset. The aim of this study is to evaluate the transferability of three published OBIA landslide detection methods for semi-automated landslide detection in the Himalaya mountainous region of northern Pakistan. A SPOT-6 multi-spectral image with Advanced Land Observing Satellite (ALOS) Phased Array type L-band Synthetic Aperture Radar (PALSAR) Digital Elevation Model (DEM) derivatives, i.e. slope, aspect, hill-shade, relief, elevation and stream network are used for landslide detection using eCognition developer software. The three published methods scale parameters for image segmentation and parameter thresholds are evaluated first. It is observed that the aforementioned methods are not directly applicable to our study area and remote sensing datasets. Therefore, an alternate (proposed) method is developed for semi-automated landslide detection. Accuracy assessment of the selected methods and proposed method is assessed by Precision, Recall and F1 measures. Using the proposed method, a total of 357 landslides are detected with 91.46% Precision, 93.31% Recall and 92.38% F1 measure accuracy.

KW - ITC-ISI-JOURNAL-ARTICLE

KW - UT-Hybrid-D

UR - https://ezproxy2.utwente.nl/login?url=https://doi.org/10.1080/01431161.2019.1701725

UR - https://ezproxy2.utwente.nl/login?url=https://library.itc.utwente.nl/login/2020/isi/vanderwerff_tra.pdf

U2 - 10.1080/01431161.2019.1701725

DO - 10.1080/01431161.2019.1701725

M3 - Article

AN - SCOPUS:85077251845

VL - 41

SP - 3390

EP - 3410

JO - International journal of remote sensing

JF - International journal of remote sensing

SN - 0143-1161

IS - 9

ER -