QUANTITATIVE ESTIMATION OF DESERTIFICATION DEGREE BASED ON PCA 
FUSION AND SVM USING CBERS-02B 
HE Qisheng 3 , Li Guoping b ,CAO Chunxiang 3 *,ZHANG Hao a ,BAO Yunfei 3 , CHANG Chaoyi 3 ,LI Xiaowen c 
a State Key Laboratory of Remote Sensing Science, Jointly Sponsored by the Institute of Remote Sensing Applications 
of Chinese Academy of Sciences and Beijing Normal University, Beijing 100101, China(hqsfei999@ 163.com, 
cao413@irsa.ac.cn, zhangh@irsa.ac.cn,byf_sheep@ 163.com, qiidee@vip.sina.com) 
b School of Earth and Space Sciences, Peking University, Beijing 100871, China - lgp@cnsa.gov.cn 
c State Key Laboratory of Remote Sensing Science, Jointly Sponsored by Beijing Normal University and the Institute of 
Remote Sensing Applications of Chinese Academy of Sciences, Beijing 100875, China, lix@bnu.edu.cn 
KEY WORDS: support vector machine (SVM); data fusion CBERS-02B soil desertification information, PCA 
ABSTRACT: 
Soil Desertification has severely threatened inner stability and sustainable development in arid areas, so extracting soil 
desertification information based on remote sensing and mastering its spatial distribution are of important practical significance. In 
this article, taking the Mu Us sand land as a case study, utilizing CBERS-02B including CCD data and HR data, the approach of 
effective remote-sensing information extraction for soil desertification in arid areas based on principal component fusion and SVM 
is discussed. Firstly, two dimensional discrete stationary wavelet transform combined average filter processing is presented to 
improve the image quality. Then principal component fusion and SVM classification were used to extract soil desertification 
information. Finally, the kernel type of radial basis function was selected and comparing the classification results in SVM 
classification with maximum likelihood classification and minimum distance classification qualitatively and quantitatively in terms 
of classification accuracy. The results suggest that the method is effective and the precision of this approach is very high, so it is an 
effective method for monitoring soil desertification changes utilizing CBERS-02B data in arid area. 
1. INTRODUCTION 
Desertification is the label for land degradation in arid, semi- 
arid, and dry sub-humid areas, collectively called dry-lands, and 
represents one of the most threatening environmental hazards 
due to the large amount of people and land at risk. Soil 
desertification has severely threatened inner stability and 
sustainable development in arid areas. Sandy desertification is 
one of the main forms of land degradation in China, especially 
in northern China (Wang T and Zhu Z D), which has been 
expanding since the 1950’s and has exerted severe impact on 
regional socio-economic development and environmental 
security (Wang T, et al). Harsh physiographic conditions 
(sparse vegetation coverage, sandy soil and water deficiency), 
irrational land-use practices, and population augmentation are 
regarded as the forces triggering sandy desertification (Zhu Z D, 
Chen Y F). Therefore, sandy desertification assessment and 
monitoring are always concerning to researchers, the public, 
and policy-makers. Remote sensing data and techniques have 
been widely used to map soil desertification areas (Liu H J, et 
al, Wang X Q, et al, Guo J, et al, Liu W J, et al). 
2. AREA DESCRIPTION AND DATA SET 
2.1 Study area 
Mu Us sand land is one of twelve large desert areas in the 
southeast of China's desert areas, at latitude 37°27.5 
39°22.5'N, longitude 107°20 111°30'E. The climate of this 
area is located in arid, semi-arid transition zone, most of the 
area is in temperate semi-arid area, and has a fragile natural, 
economic and social complex ecosystem. For a long time, due 
to lack of ecological protection awareness and blind pursuit of 
short profit, a vicious cyle of “population growth - lower 
standard of living - over-development - ecological degradation - 
low productivity - over-development” damaged the ecological 
environment. The location of the study area is shown in Fig.l. 
2.2 Data set 
The CBERS-02B satellite was set up by China on September 19, 
2007, with the highest resolution HR image by 2.36m and the 
middle resolution CCD image. The CBERS-02B HR data is 
taken almost simultaneously with multi-spectral CCD data 
Corresponding author. 
299