Evaluation of ForestPA for VHR RS image classification using spectral and superpixel-guided morphological profiles

Alim Samat, Sicong Liu, C. Persello, Erzhu Li, Zelang Miao, Jilili Abuduwaili

Research output: Contribution to journalArticleAcademicpeer-review

5 Downloads (Pure)

Abstract

In very high resolution (VHR) remote sensing (RS) classification tasks, conventional pixel-based contextual information extraction methods such as morphological profiles (MPs), extended MPs (EMPs) and MPs with partial reconstruction (MPPR) with limited numbers, sizes and shapes of structural elements (SEs) cannot perfectly match all sizes and shapes of the objects in an image. To overcome such limitation, we introduce novel spatial feature extractors, namely, the superpixel-guided morphological profiles (SPMPs), where the superpixels are used as SEs in opening by reconstruction and closing by reconstruction operations. Moreover, to avoid possible side effects from unusual maximum and minimum values within superpixels, the mean pixel value of superpixels is adopted (SPMPsM). Additionally, new decision forest based on penalizing the attributes in previous trees, the ForestPA is introduced and evaluated through a comparative investigation on three VHR multi-/hyperspectral RS image classification tasks. Support vector machine and benchmark ensemble classifiers, including bagging, AdaBoost, MultiBoost, ExtraTrees, Random Forest and Rotation Forest, are adopted. The experimental results confirm the effectiveness and superior performances of the proposed SPMPs and SPMPsM relative to those of the MPs and MPPR. Moreover, ForestPA outperforms only bagging and is not suitable for learning from large numbers of samples with high dimensionality from the computational efficiency and classification accuracy perspective.

Original languageEnglish
Pages (from-to)107-121
Number of pages15
JournalEuropean journal of remote sensing
Volume52
Issue number1
DOIs
Publication statusPublished - 1 Jan 2019

Fingerprint

Image classification
Remote Sensing Image
Image Classification
image classification
Remote sensing
High Resolution
Pixels
remote sensing
Adaptive boosting
Evaluation
Computational efficiency
Support vector machines
pixel
Bagging
Classifiers
extraction method
Pixel
Hyperspectral Remote Sensing
Ensemble Classifier
Partial

Keywords

  • ForestPA
  • image classification
  • MPPR
  • MPs
  • superpixel
  • superpixel-guided morphological profiles
  • VHR images
  • ITC-ISI-JOURNAL-ARTICLE
  • ITC-GOLD

Cite this

Samat, Alim ; Liu, Sicong ; Persello, C. ; Li, Erzhu ; Miao, Zelang ; Abuduwaili, Jilili. / Evaluation of ForestPA for VHR RS image classification using spectral and superpixel-guided morphological profiles. In: European journal of remote sensing. 2019 ; Vol. 52, No. 1. pp. 107-121.
@article{159c431e68a748f186ce8412578b341a,
title = "Evaluation of ForestPA for VHR RS image classification using spectral and superpixel-guided morphological profiles",
abstract = "In very high resolution (VHR) remote sensing (RS) classification tasks, conventional pixel-based contextual information extraction methods such as morphological profiles (MPs), extended MPs (EMPs) and MPs with partial reconstruction (MPPR) with limited numbers, sizes and shapes of structural elements (SEs) cannot perfectly match all sizes and shapes of the objects in an image. To overcome such limitation, we introduce novel spatial feature extractors, namely, the superpixel-guided morphological profiles (SPMPs), where the superpixels are used as SEs in opening by reconstruction and closing by reconstruction operations. Moreover, to avoid possible side effects from unusual maximum and minimum values within superpixels, the mean pixel value of superpixels is adopted (SPMPsM). Additionally, new decision forest based on penalizing the attributes in previous trees, the ForestPA is introduced and evaluated through a comparative investigation on three VHR multi-/hyperspectral RS image classification tasks. Support vector machine and benchmark ensemble classifiers, including bagging, AdaBoost, MultiBoost, ExtraTrees, Random Forest and Rotation Forest, are adopted. The experimental results confirm the effectiveness and superior performances of the proposed SPMPs and SPMPsM relative to those of the MPs and MPPR. Moreover, ForestPA outperforms only bagging and is not suitable for learning from large numbers of samples with high dimensionality from the computational efficiency and classification accuracy perspective.",
keywords = "ForestPA, image classification, MPPR, MPs, superpixel, superpixel-guided morphological profiles, VHR images, ITC-ISI-JOURNAL-ARTICLE, ITC-GOLD",
author = "Alim Samat and Sicong Liu and C. Persello and Erzhu Li and Zelang Miao and Jilili Abuduwaili",
year = "2019",
month = "1",
day = "1",
doi = "10.1080/22797254.2019.1565418",
language = "English",
volume = "52",
pages = "107--121",
journal = "European journal of remote sensing",
issn = "2279-7254",
publisher = "Taylor & Francis",
number = "1",

}

Evaluation of ForestPA for VHR RS image classification using spectral and superpixel-guided morphological profiles. / Samat, Alim; Liu, Sicong; Persello, C.; Li, Erzhu; Miao, Zelang; Abuduwaili, Jilili.

In: European journal of remote sensing, Vol. 52, No. 1, 01.01.2019, p. 107-121.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - Evaluation of ForestPA for VHR RS image classification using spectral and superpixel-guided morphological profiles

AU - Samat, Alim

AU - Liu, Sicong

AU - Persello, C.

AU - Li, Erzhu

AU - Miao, Zelang

AU - Abuduwaili, Jilili

PY - 2019/1/1

Y1 - 2019/1/1

N2 - In very high resolution (VHR) remote sensing (RS) classification tasks, conventional pixel-based contextual information extraction methods such as morphological profiles (MPs), extended MPs (EMPs) and MPs with partial reconstruction (MPPR) with limited numbers, sizes and shapes of structural elements (SEs) cannot perfectly match all sizes and shapes of the objects in an image. To overcome such limitation, we introduce novel spatial feature extractors, namely, the superpixel-guided morphological profiles (SPMPs), where the superpixels are used as SEs in opening by reconstruction and closing by reconstruction operations. Moreover, to avoid possible side effects from unusual maximum and minimum values within superpixels, the mean pixel value of superpixels is adopted (SPMPsM). Additionally, new decision forest based on penalizing the attributes in previous trees, the ForestPA is introduced and evaluated through a comparative investigation on three VHR multi-/hyperspectral RS image classification tasks. Support vector machine and benchmark ensemble classifiers, including bagging, AdaBoost, MultiBoost, ExtraTrees, Random Forest and Rotation Forest, are adopted. The experimental results confirm the effectiveness and superior performances of the proposed SPMPs and SPMPsM relative to those of the MPs and MPPR. Moreover, ForestPA outperforms only bagging and is not suitable for learning from large numbers of samples with high dimensionality from the computational efficiency and classification accuracy perspective.

AB - In very high resolution (VHR) remote sensing (RS) classification tasks, conventional pixel-based contextual information extraction methods such as morphological profiles (MPs), extended MPs (EMPs) and MPs with partial reconstruction (MPPR) with limited numbers, sizes and shapes of structural elements (SEs) cannot perfectly match all sizes and shapes of the objects in an image. To overcome such limitation, we introduce novel spatial feature extractors, namely, the superpixel-guided morphological profiles (SPMPs), where the superpixels are used as SEs in opening by reconstruction and closing by reconstruction operations. Moreover, to avoid possible side effects from unusual maximum and minimum values within superpixels, the mean pixel value of superpixels is adopted (SPMPsM). Additionally, new decision forest based on penalizing the attributes in previous trees, the ForestPA is introduced and evaluated through a comparative investigation on three VHR multi-/hyperspectral RS image classification tasks. Support vector machine and benchmark ensemble classifiers, including bagging, AdaBoost, MultiBoost, ExtraTrees, Random Forest and Rotation Forest, are adopted. The experimental results confirm the effectiveness and superior performances of the proposed SPMPs and SPMPsM relative to those of the MPs and MPPR. Moreover, ForestPA outperforms only bagging and is not suitable for learning from large numbers of samples with high dimensionality from the computational efficiency and classification accuracy perspective.

KW - ForestPA

KW - image classification

KW - MPPR

KW - MPs

KW - superpixel

KW - superpixel-guided morphological profiles

KW - VHR images

KW - ITC-ISI-JOURNAL-ARTICLE

KW - ITC-GOLD

UR - https://ezproxy2.utwente.nl/login?url=https://library.itc.utwente.nl/login/2019/isi/persello_eva.pdf

U2 - 10.1080/22797254.2019.1565418

DO - 10.1080/22797254.2019.1565418

M3 - Article

VL - 52

SP - 107

EP - 121

JO - European journal of remote sensing

JF - European journal of remote sensing

SN - 2279-7254

IS - 1

ER -