Generalized differential morphological profiles for remote sensing image classification

Xin Huang, Xiaopeng Han, Liangpei Zhang, Jianya Gong, Wenzhi Liao, Jon Atli Benediktsson, Jocelyn Chanussot

Research output: Contribution to journalArticle

17 Citations (Scopus)

Abstract

Differential morphological profiles (DMPs) are widely used for the spatial/structural feature extraction and classification of remote sensing images. They can be regarded as the shape spectrum, depicting the response of the image structures related to different scales and sizes of the structural elements (SEs). DMPs are defined as the difference of morphological profiles (MPs) between consecutive scales. However, traditional DMPs can ignore discriminative information for features that are across the scales in the profiles. To solve this problem, we propose scale-span differential profiles, i.e., generalized DMPs (GDMPs), to obtain the entire differential profiles. GDMPs can describe the complete shape spectrum and measure the difference between arbitrary scales, which is more appropriate for representing the multiscale characteristics and complex landscapes of remote sensing image scenes. Subsequently, the random forest (RF) classifier is applied to interpret GDMPs considering its robustness for high-dimensional data and ability of evaluating the importance of variables. Meanwhile, the RF "out-of-bag" error can be used to quantify the importance of each channel of GDMPs and select the most discriminative information in the entire profiles. Experiments conducted on three well-known hyperspectral data sets as well as an additional World View-2 data are used to validate the effectiveness of GDMPs compared to the traditional DMPs. The results are promising as GDMPs can significantly outperform the traditional one, as it is capable of adequately exploring the multiscale morphological information.
Original languageEnglish
Pages (from-to)1736-1751
Number of pages16
JournalIEEE Journal of Selected Topics in Earth Observation and Remote Sensing
Volume9
Issue number4
DOIs
Publication statusPublished - 26 Feb 2016

Fingerprint

Image classification
image classification
Remote sensing
remote sensing
Feature extraction
Classifiers
Experiments
experiment

Keywords

  • random forest (RF)
  • morphological profiles
  • feature selection
  • classification.
  • feature extraction
  • supervised feature extraction
  • local binary patterns
  • urban areas
  • spatial classification
  • attribute profiles
  • hyperspectral data

Cite this

Huang, Xin ; Han, Xiaopeng ; Zhang, Liangpei ; Gong, Jianya ; Liao, Wenzhi ; Benediktsson, Jon Atli ; Chanussot, Jocelyn. / Generalized differential morphological profiles for remote sensing image classification. In: IEEE Journal of Selected Topics in Earth Observation and Remote Sensing. 2016 ; Vol. 9, No. 4. pp. 1736-1751.
@article{ca8cd88f32544609a42c370edf6dc6f0,
title = "Generalized differential morphological profiles for remote sensing image classification",
abstract = "Differential morphological profiles (DMPs) are widely used for the spatial/structural feature extraction and classification of remote sensing images. They can be regarded as the shape spectrum, depicting the response of the image structures related to different scales and sizes of the structural elements (SEs). DMPs are defined as the difference of morphological profiles (MPs) between consecutive scales. However, traditional DMPs can ignore discriminative information for features that are across the scales in the profiles. To solve this problem, we propose scale-span differential profiles, i.e., generalized DMPs (GDMPs), to obtain the entire differential profiles. GDMPs can describe the complete shape spectrum and measure the difference between arbitrary scales, which is more appropriate for representing the multiscale characteristics and complex landscapes of remote sensing image scenes. Subsequently, the random forest (RF) classifier is applied to interpret GDMPs considering its robustness for high-dimensional data and ability of evaluating the importance of variables. Meanwhile, the RF {"}out-of-bag{"} error can be used to quantify the importance of each channel of GDMPs and select the most discriminative information in the entire profiles. Experiments conducted on three well-known hyperspectral data sets as well as an additional World View-2 data are used to validate the effectiveness of GDMPs compared to the traditional DMPs. The results are promising as GDMPs can significantly outperform the traditional one, as it is capable of adequately exploring the multiscale morphological information.",
keywords = "random forest (RF), morphological profiles, feature selection, classification., feature extraction, supervised feature extraction, local binary patterns, urban areas, spatial classification, attribute profiles, hyperspectral data",
author = "Xin Huang and Xiaopeng Han and Liangpei Zhang and Jianya Gong and Wenzhi Liao and Benediktsson, {Jon Atli} and Jocelyn Chanussot",
year = "2016",
month = "2",
day = "26",
doi = "10.1109/JSTARS.2016.2524586",
language = "English",
volume = "9",
pages = "1736--1751",
journal = "IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing",
issn = "1939-1404",
publisher = "IEEE",
number = "4",

}

Generalized differential morphological profiles for remote sensing image classification. / Huang, Xin; Han, Xiaopeng; Zhang, Liangpei; Gong, Jianya; Liao, Wenzhi; Benediktsson, Jon Atli; Chanussot, Jocelyn.

In: IEEE Journal of Selected Topics in Earth Observation and Remote Sensing, Vol. 9, No. 4, 26.02.2016, p. 1736-1751.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Generalized differential morphological profiles for remote sensing image classification

AU - Huang, Xin

AU - Han, Xiaopeng

AU - Zhang, Liangpei

AU - Gong, Jianya

AU - Liao, Wenzhi

AU - Benediktsson, Jon Atli

AU - Chanussot, Jocelyn

PY - 2016/2/26

Y1 - 2016/2/26

N2 - Differential morphological profiles (DMPs) are widely used for the spatial/structural feature extraction and classification of remote sensing images. They can be regarded as the shape spectrum, depicting the response of the image structures related to different scales and sizes of the structural elements (SEs). DMPs are defined as the difference of morphological profiles (MPs) between consecutive scales. However, traditional DMPs can ignore discriminative information for features that are across the scales in the profiles. To solve this problem, we propose scale-span differential profiles, i.e., generalized DMPs (GDMPs), to obtain the entire differential profiles. GDMPs can describe the complete shape spectrum and measure the difference between arbitrary scales, which is more appropriate for representing the multiscale characteristics and complex landscapes of remote sensing image scenes. Subsequently, the random forest (RF) classifier is applied to interpret GDMPs considering its robustness for high-dimensional data and ability of evaluating the importance of variables. Meanwhile, the RF "out-of-bag" error can be used to quantify the importance of each channel of GDMPs and select the most discriminative information in the entire profiles. Experiments conducted on three well-known hyperspectral data sets as well as an additional World View-2 data are used to validate the effectiveness of GDMPs compared to the traditional DMPs. The results are promising as GDMPs can significantly outperform the traditional one, as it is capable of adequately exploring the multiscale morphological information.

AB - Differential morphological profiles (DMPs) are widely used for the spatial/structural feature extraction and classification of remote sensing images. They can be regarded as the shape spectrum, depicting the response of the image structures related to different scales and sizes of the structural elements (SEs). DMPs are defined as the difference of morphological profiles (MPs) between consecutive scales. However, traditional DMPs can ignore discriminative information for features that are across the scales in the profiles. To solve this problem, we propose scale-span differential profiles, i.e., generalized DMPs (GDMPs), to obtain the entire differential profiles. GDMPs can describe the complete shape spectrum and measure the difference between arbitrary scales, which is more appropriate for representing the multiscale characteristics and complex landscapes of remote sensing image scenes. Subsequently, the random forest (RF) classifier is applied to interpret GDMPs considering its robustness for high-dimensional data and ability of evaluating the importance of variables. Meanwhile, the RF "out-of-bag" error can be used to quantify the importance of each channel of GDMPs and select the most discriminative information in the entire profiles. Experiments conducted on three well-known hyperspectral data sets as well as an additional World View-2 data are used to validate the effectiveness of GDMPs compared to the traditional DMPs. The results are promising as GDMPs can significantly outperform the traditional one, as it is capable of adequately exploring the multiscale morphological information.

KW - random forest (RF)

KW - morphological profiles

KW - feature selection

KW - classification.

KW - feature extraction

KW - supervised feature extraction

KW - local binary patterns

KW - urban areas

KW - spatial classification

KW - attribute profiles

KW - hyperspectral data

UR - http://hdl.handle.net/1854/LU-7064981

U2 - 10.1109/JSTARS.2016.2524586

DO - 10.1109/JSTARS.2016.2524586

M3 - Article

VL - 9

SP - 1736

EP - 1751

JO - IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing

JF - IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing

SN - 1939-1404

IS - 4

ER -