A Random Subspace based classification for temporal data of satellite Images

Author Name(s): K.Radhika, S.Varadarajan
Author Email: krkvarma15@gmail.com

Abstract

Change detection techniques are playing major role in order to know the changes in different situations. Using these methods, we can know the changes that are occurred or occurring on earth due to climatically changes or degradation of certain areas. Change detection will help us to observe the different areas on earth remotely. The correct changes can be obtained via proper classification methods. Generally a single classifier may not meet the current requirement. So that ensemble classifiers are required. Wrong classification results incorrect changes. In this paper, ensemble subspace classifier is discussed to get accurate classification thereby accurate changes can be obtained. The changed area is calculated based on the comparison of the number of pixels in each class of two temporal images. Here we have shown the changed pixels in each class and also changed areas in each and every class.

Introduction

Recently, changed detection technology of remote sensing image is developing rapidly, and varietyof change detection algorithms are proposed by some researchers. The changes occurring in our surroundings and environment need to be observed. It is possible with satellite images. It is one of the major branches of Remote Sensing image processing. To apply this there are several applications in our common life like, manmade target detection, urban planning, land-cover database updating, disaster management and the development of human civilization. So that there is a great need to detect the changes. With the help of sensor technology multi-temporal satellite image are available with low resolution and high resolution so that monitoring of earth surface time-to-time is possible [1-3].

Change detection will identify certain changes in the images of same area taken at different time instants. The under lying information can be extracted to know the changes easily over large areas on the earth surface. Change detection techniques are based on multiband, multi-temporal or single-phase images using Image processing methods. The analysis is carried out based on pixel level, feature level and target level. Classification can be supervised or unsupervised.

Generally the satellite images consists some complex backgrounds which will reduce the efficiency of change detection methods [4-5]. Still the research is going on different sophisticated kind of change detection models. The major procedure to find the changes are, obtaining difference image and following correct and exact classifiers. Two basic procedures to get difference image are, one is subtraction operation and another is ratio operation. The subtraction of the pixels in the bi-temporal images, the changes can be observed. The other methods are mean-filter, log-ratio and mean-ratio etc.Only one difference image is produced majorly in all change detection methods [6-7].

Conclusion

In this paper the discussions are carried-out for two set of images on ensemble techniques for change detection. 200 points were considered with four classes: water, green land, barren land and agriculture for validation of output. Finally, pixels of each class were compared in the two images and also calculated the corresponding area. This analysis gave good change detection based on ensemble method. This can be used in disaster management. Even we have verified this procedure with support vector machine, k nearest neighbors, random forest and maximum likelihood procedures. This change detection can be useful in forest deformation, crop monitoring, urban area development, etc.

References

[1] LStanistaw. ObjectOriented Classification of Landsat ETM+ Satellite Image. Journal of Water and   Land Development, Vol. 20, No. 10, pp. 91-106, 2006.

[2] Y Zhong, L Zhang. Remote sensing image subpixel mapping based on adaptive differential evolution. IEEE Trans. Syst., Man.vol. 42, no. 5, pp. 1306–1329, Oct. 2012.

[3] J Epstein, Payne K, Kramer E. Techniques for mapping suburban sprawl. Photogrammetric Engineering and Remote Sensing. Vol.63, No.9, pp.913-918, 2009.

[4] SHemissi, I R Farah, K Ettabaa, BSolaiman. Multi-spectrotemporal analysis of hyperspectralimagery based on 3-D spectral modeling and multilinear algebra. IEEE Trans. Geosci.Remote Sens., vol.51, no. 1, pp. 199–216, Jan. 2013.

[5] F De Morsier, D Tuia, M Borgeaud, V Gass, J P Thiran. Semisupervised novelty detection using SVM entire solution path. IEEE Trans. Geosci. Remote Sens., vol. 51, no. 4, pp. 1939–1950, Apr. 2013.

[6] T. Celik. Change detection in satellite images using a genetic algorithm approach. IEEE Geosci. Remote Sens. Lett., vol. 7, no. 2, pp. 386–390, Apr. 2010.

[7] K Chen, Z Zhou, C Huo, X Sun, K Fu. A semisupervised context-sensitive change detection  technique via Gaussian process. IEEE Geosci. Remote Sens. Lett., vol. 10, no. 2, pp. 236–240, Mar. 2013.

[8] F Bovolo, SMarchesi, L Bruzzone. A framework for automatic and unsupervised detection of multiple changes in multitemporal images. IEEE Trans. Geosci. Remote Sens., vol. 50, no. 6, pp. 2196-2212, Jun. 2012.

[9] K Radhika, S Varadarajan. A Tutorial on Classification of Remote Sensing Data. International Research Journal of Engineering and Technology, Volume 3, Issue 8, 881 – 885, August. 2015.

[10] H Li, G Xiao, T Xia, Y Y. Tang, L Li.Hyperspectral image classification using functional data  Analysis. IEEE Trans. Cybern., vol. 44, no. 9, pp. 1544–1555, Sep. 2014.

[11] Wei Li, Miao Lu,Xiuwan Chen. Automatic change detection of urban Land-cover based on SVM classification. IGARSS,IEEE,2015.

[12] Xiaoqiang Lu, Yuan Yuan, XiangtaoZheng. Joint dictionary learning for multispectral change detection. IEEE transactions on cybernetics,2016.

709 total views, no views today

Download PDF File

About the author: admin