[논문]Application of Random Forests to Assessment of Importance of Variables in Multi-sensor Data Fusion for Land-cover Classification

Park No-Wook; Chi kwang-Hoon

doi:10.7780/kjrs.2006.22.3.211

[국내논문] Application of Random Forests to Assessment of Importance of Variables in Multi-sensor Data Fusion for Land-cover Classification 원문보기

대한원격탐사학회지 = Korean journal of remote sensing, v.22 no.3, 2006년, pp.211 - 219

Park No-Wook (Geoscience Information Center, Korea Institute of Geoscience and Mineral Resources) , Chi kwang-Hoon (Geoscience Information Center, Korea Institute of Geoscience and Mineral Resources)

Abstract ▼ AI-Helper

A random forests classifier is applied to multi-sensor data fusion for supervised land-cover classification in order to account for the importance of variable. The random forests approach is a non-parametric ensemble classifier based on CART-like trees. The distinguished feature is that the importance of variable can be estimated by randomly permuting the variable of interest in all the out-of-bag samples for each classifier. Two different multi-sensor data sets for supervised classification were used to illustrate the applicability of random forests: one with optical and polarimetric SAR data and the other with multi-temporal Radarsat-l and ENVISAT ASAR data sets. From the experimental results, the random forests approach could extract important variables or bands for land-cover discrimination and showed reasonably good performance in terms of classification accuracy.

주제어

AI 본문요약
AI-Helper

* AI 자동 식별 결과로 적합하지 않은 문장이 있을 수 있으니, 이용에 유의하시기 바랍니다.

문제 정의

We also thank Professor Hoonyol Lee of Kangwon National University for his insightful comments on coherence imagery generation. This work was supported by the public application research of satellite data.

가설 설정

The importance of variables is shown in Fig. 1. All importance values were normalized to express the relative ones with respect to the maximum. If the average increased out-of-bag error value is large, the variable can be considered more important.
Though the paddy and dry fields generally show higher temporal variability than those of other classes, intensity or property of temporal variability between those two cultivated areas may be quite different due to the different cultivation mechanism during the plant growth cycle. In this experiment, the paddy and dry fields were considered as the separated two landcover classes. The standard deviation values were used as temporal variability factors and the paddy and dry fields were considered as independent two landcover classes.

제안 방법

To account for relative reliability or importance, this paper applies random forests to multi-sensor data fusion for land-cover classification. The random forests approach is one of ensemble methods and a non-parametric one and thus it can be effectively applied to multi-sensor data fusion.
(1996), the probabilistic neural network(PNN) and k-nm method were applied. For accuracy statistics, overall accuracy, user's accuracy, average accuracy and Kappa statistic were computed by constructing a confusion matrix (Table 1). The improvements for the stubble and carrots classes were significant but the performance for the bare soil class was poorer than PNN and in algorithms.
Since the RadarsaM and ENVISAT ASAR data were acquired from different orbits, they showed the different imaging geometry. To reduce the effect of topography on the backscattering coefficient, geocoding with DEM extracted from a 1:25,000 scale digital topographic map of the study area was done. The lay over and shadow zones extracted during geocoding with DEM were masked out throughout the data processing for classification.
Before classification, first, a feature extraction step was applied by considering the scattering properties of multi-temporal SAR data. This study considered three features including average backscattering coefficient, temporal variability and long-term coherence like previous researches (Strozzi et al., 2000; Park et al., 2005). From the average backscattering coefficient, water and urban areas can be discriminated due to their very low and high backscattering coefficient, respectively.
The superiority of the features from Radarsat-1 data to those from ENVISAT ASAR data however does not explain that HH polarization data always give us more information than the W polarization data in an agricultural area. It should be noted that the results are the combined results of various parameters such as acquisition dates, incidence angle, mode, polarization state, etc. The coherence values of both Radarsat-1 and ENVISAT ASAR data sets have little contributions to discrimination of the land-cover classes in the study area.

대상 데이터

The second experiment was carried out in the Yedang plain, Korea. The data sets considered in flie second experiment include multi-temporal C-band Radarsat-1 data (HH polarization) and ENVISAT ASAR data (VV polarization) (Table 2). The Radarsat-1 data acquired from April, 2005 to October, 2005 were used and ENVISAT ASAR data span a whole year from October, 2004 to October, 2005.
The provision of the multi-sensor data set (grss- dfc-0006) by IEEE GRSS Data Fusion Technical Committee is gratefully acknowledged. Three data among the ENVISAT AS AR data sets used in this study were kindly provided by the Korea Aerospace Research Institute (KARI). We also thank Professor Hoonyol Lee of Kangwon National University for his insightful comments on coherence imagery generation.

이론/모형

The multi-sensor data set (grss-dfc-0006) used in the experiment for supervised land-cover classification was provided by the IEEE GRSS Data Fusion Committee (http://www.dfc-grss.org). It includes airborne Thematic Mapper Scanner data with 6 channels and NASA JPL AirSAR data with 9 channels in the C-, L- and P-bands and HH, HV and VV polarizations.

성능/효과

Furthermore, the importance of variable for each land-cover class was also computed and is shown in Fig. 2. Like the case of the whole importance values, the contribution of optical data is dominant than those of SAR data. L- band polarimetric data however played a major role in the discrimination of potatoes class.
For accuracy statistics, overall accuracy, user's accuracy, average accuracy and Kappa statistic were computed by constructing a confusion matrix (Table 1). The improvements for the stubble and carrots classes were significant but the performance for the bare soil class was poorer than PNN and in algorithms. Overall, random forests show similar or improved classification performances.

참고문헌 (14)

Breiman, L., 2001. Random forests, Machine Learning, 45(1): 5-32
Breiman, L., J. H. Friedman, R. A. Olshen, and C. J. Stone, 1984. Classification and regression trees, Wadsworth Inc
Bruzzone, L., M. Marconcini, U. Wegmuller, and A. Wiesmann, 2004. An advanced system for the automatic classification of multitemporal SAR images, IEEE Transactions on Geoscience and Remote Sensing, 42(6): 1321-1334

상세보기
Gislason, P. O., J. A. Benediktsson, and J. R. Sveinsson, 2004. Random forest classification of multisource remote sensing and geographic data, Proceedings of IGARSS 2004
Ham, J., Y. Chen, M. M. Crawford, and J. Ghosh, 2005. Investigation of the random forest framework for classification of hyperspectral data, IEEE Transactions on Geoscience and Remote Sensing, 43(3): 492-501

상세보기
Joelsson, S. R., J. A. Benediktsson, and J. R. Sveinsson, 2005. Random forest classifiers for hyperspectral data, Proceedings of IGARSS 2005
Le Hegarat-Mascle, S., I. Bloch, and D. Vidal-Madjar, 1997. Application of Dempster-Shafer evidence theory to unsupervised classification in multisource remote sensing, IEEE Transactions on Geoscience and Remote Sensing, 35(4): 1018-1031

상세보기
Park, N.-W., 2004. Multi-source spatial data fusion with geostatistical uncertainty assessment: applications to landslide susceptibility analysis and land-cover classification, Ph. D. Thesis, Seoul National University
Park, N.-W., H. Lee, and K.-H. Chi, 2005. Feature extraction and fusion for land-cover discrimination with multi-temporal SAR data, Korean Journal of Remote Sensing, 21(2): 145-162
Serpico, S. B., L. Bruzzone, and F. Roli, 1996. An experimental comparison of neural and statistical non-parametric algorithms for supervised classification of remote-sensing images, Pattern Recognition Letters, 17(13): 1331-1341

상세보기
Solberg, A. H. S., T. Taxt, and A. K. Jain, 1996. A Markov random field model for classification of multisource satellite imagery, IEEE Transactions on Geoscience and Remote Sensing, 34(1): 100-113

상세보기
Solaiman, B., L. E. Pierce, and F. T. Ulaby, 1999. Multisensor data fusion using fuzzy concepts: application to land-cover classification using ERS-1/JERS-1 SAR composites, IEEE Transactions on Geoscience and Remote Sensing, 37(3): 1316-1326

상세보기
Strozzi, T., P. B. G. Dammert, U. Wegmuller, J. M. Martinez, A. Beaudoin, J. Askne, and M. Hallikainen, 2000. Landuse mapping with ERS SAR interferometry, IEEE Transactions on Geoscience and Remote Sensing, 38(2): 766-775

상세보기
Wegmuller, U., T. Strozzi, A. Wiesmann, and C. Werner, 2003. ENVISAT ASAR for land cover information, Proceedings of IGARSS 2003

내보내기 구분	파일저장 인쇄 메일전송
구성항목	기본정보 상세정보 관리번호, 논문명, 저널/프로시딩명, 저자 , 발행년, 권, 호, 시작페이지, 끝페이지, 발행기관 관리번호, 논문명, 대등논문명, 저자 , 저널/프로시딩명, 발행기관, 발행년, 발행언어, 권, 호, 시작페이지, 끝페이지, ISBN, ISSN, 주제분야, 키워드, 초록(한글), 초록(영문), 저자(소속기관)
저장형식	Text(ASCII format) Excel format RefWorks Direct Export RIS format (for Reference Manager, ProCite, EndNote), Scholar's Aids, Mendeley
메일정보	받는사람 (필수) @ 보내는사람 (선택) @ 제목 내용 KISTI 검색결과 이메일 서비스
안내	총 건의 자료가 검색되었습니다. 다운받으실 자료의 인덱스를 입력하세요. (1-10,000) 검색결과의 순서대로 최대 10,000건 까지 다운로드가 가능합니다. 데이타가 많을 경우 속도가 느려질 수 있습니다.(최대 2~3분 소요) 다운로드 파일은 UTF-8 형태로 저장됩니다. 파일의 내용이 제대로 보이지 않을실 때는 웹브라우저 상단의 보기 -> 인코딩 -> 자동선택 여부를 확인하십시오. ~ Text(ASCII format) Excel format

연합인증

[국내논문] Application of Random Forests to Assessment of Importance of Variables in Multi-sensor Data Fusion for Land-cover Classification 원문보기

Abstract ▼ AI-Helper

주제어

AI 본문요약
AI-Helper

문제 정의

가설 설정

제안 방법

대상 데이터

이론/모형

성능/효과

참고문헌 (14)

이 논문을 인용한 문헌

저자의 다른 논문 :

관련 콘텐츠

원문 보기

원문 URL 링크

오픈액세스(OA) 유형

이 논문과 함께 이용한 콘텐츠

AI-Helper ※ AI-Helper는 오픈소스 모델을 사용합니다.

선택된 텍스트

연합인증

[국내논문] Application of Random Forests to Assessment of Importance of Variables in Multi-sensor Data Fusion for Land-cover Classification 원문보기

Abstract ▼ AI-Helper

주제어

AI 본문요약 엑셀 다운로드 AI-Helper

문제 정의

가설 설정

제안 방법

대상 데이터

이론/모형

성능/효과

참고문헌 (14)

이 논문을 인용한 문헌

저자의 다른 논문 :

박노욱 (19) 지광훈 (31)

관련 콘텐츠

원문 보기

원문 URL 링크

오픈액세스(OA) 유형

이 논문과 함께 이용한 콘텐츠

AI-Helper ※ AI-Helper는 오픈소스 모델을 사용합니다.

선택된 텍스트

AI 본문요약
AI-Helper