以多時期與PCA+NDVI法改善地物分類之正確性與完整性

Using Multi-temporal and PCA+NDVI to Improve the Accuracy and Integrity of Land Cover Classification

10.6234/JGR.2012.57.03

張國楨(Kuo-chen Chang)；田應平(Ying-Ping Tian)；施孝謙(Hsiao-Chien Shih)

地物分類 ； 多時期影像 ； PCA ； 非監督式分類 ； 完整性 ； Land Cover Classification ； Multi-temporal Imagery ； PCA ； Unsupervised Classification ； Completeness

地理研究

57期（2012 / 11 / 01）

49 - 60

繁體中文

Satellite image analysis is one of the main methods of monitoring of environmental changes. The accuracy and credibility of the results depend on the spectral resolution of the imagery used and the spectral separability between features being monitored. If original imagery from single period is used to perform classification, the results are often affected by the noise of imagery itself and spectral similarity between different features. In this research, we proposed an improved solution for change detection that combines a Principle Component Analysis (PCA) process to remove noise and a multi-temporal process to increase spectral resolution. The study area is located on Shezi Island with imageries from 2005, 2006, 2007. The original imageries were processed with PCA to retain the first two major components which account for over 95% of total explanation. The resulting imageries were inversed by IPCA process back to 4 bands imageries. NDVI was calculated for each time period and stacked with IPCA imageries to create a new multi-temporal imagery for further unsupervised classification. The experimental results showed that the PCA process does enhance the spectral characteristic of features being monitored. An unsupervised classification process was applied to the multi-temporal imagery. The result was compared to that from 2007 and showed a significant improvement in both accuracy and completeness of the land cover classification.

Satellite image analysis is one of the main methods of monitoring of environmental changes. The accuracy and credibility of the results depend on the spectral resolution of the imagery used and the spectral separability between features being monitored. If original imagery from single period is used to perform classification, the results are often affected by the noise of imagery itself and spectral similarity between different features. In this research, we proposed an improved solution for change detection that combines a Principle Component Analysis (PCA) process to remove noise and a multi-temporal process to increase spectral resolution. The study area is located on Shezi Island with imageries from 2005, 2006, 2007. The original imageries were processed with PCA to retain the first two major components which account for over 95% of total explanation. The resulting imageries were inversed by IPCA process back to 4 bands imageries. NDVI was calculated for each time period and stacked with IPCA imageries to create a new multi-temporal imagery for further unsupervised classification. The experimental results showed that the PCA process does enhance the spectral characteristic of features being monitored. An unsupervised classification process was applied to the multi-temporal imagery. The result was compared to that from 2007 and showed a significant improvement in both accuracy and completeness of the land cover classification.

1. Pearson, K. (1901)：On Lines and Planes of Closest Fit to Systems of Points in Space. Philosophical Magazine 2 (11): 559–572
2. Hotelling, H. (1933)：Analysis of a complex of statistical variables into principal components. Journal of Educational Psychology, 24, 417-441, 498-520
3. Chen, G.,Qian ,S.E.(2008).Evaluation and comparison of dimensionality reduction methods and band selection.Canadian Journal of Remote Sensing,34(1),26-36.
4. Dechka, J.A.,Franklin, S.E.,Watmough, M.D.,Bennett, R.P.,Ingstrup , D.W.(2002).Classification of wetland habitat and vegetation communities using multi-temporal Ikonos imagery in southern Saskatchewan.Canadian Journal of Remote Sensing,28(5),679-685.
5. Du, Y.,Teillet , P. M.,Cihlar, J.(2002).Radiometric normalization of multitemporal high-resolution satellite images with quality control for land cover change detection..Remote Sensing of Environment,82,123-134.
6. Journaux ,L.,Tizon, X.,Foucherot, I.,Gouton , P.(2006).Dimensionality reduction techniques: an operational comparison on multispectral satellite images using unsupervised clustering.NORSIG,2006,242-245.
7. Lunetta, R. S.(ed.),Elvidge, C. D.(ed.)(1998).remote sensing change detection, environmental monitoring methods and applications.Chelsea:Ann Harbor Press.
8. Singh, A.,Harrison, A.(1985).Standardized principal components.International Journal of Remote Sensing,6(6),883-896.
9. Townsend, P.A.,Walsh, S.J.(2001).Remote sensing of forested wetlands: application of multitemporal and multispectral satellite imagery to determine plant community composition and structure in southeastern USA.Plant Ecology,157,129-149.
10. Wolter, P. T.,Mladenoff, D. J.,Host G. E.,Crow, T. R.(1995).Improved forest classification in the northern lake states using multi-temporal landsat imagery.Photogrammetric Engineering & Remote Sensing,61(9),1129-1143.
11. 周明中(2005)。台北，國立中央大學土木工程研究所。
12. 郭麟霂(2000)。台北，國立交通大學土木工程學系。
13. 曾露儀(2008)。以多時序多解析度之遙測資料進行植物交錯群落監測─淡水紅樹林的研究個案。第四屆亞洲空間研討會論文集，台北:
14. 馮梓旋(2007)。台北，明道管理學院環境規劃暨設計研究所。
15. 劉守恆(2002)。台北，國立成功大學地球科學研究所。
16. 蕭國鑫(1998)。台北，國立交通大學土木工程學系。
17. 羅俊宏(2004)。台北，國立中央大學通訊工程研究所。