Focused on the cloud detection task using EOS/MODIS (Earth Observation System/Moderate Resolution Imaging
Spectroradiometer) information, this paper introduced a new method for cloud detection by use of the Support Vector
Machines (SVMs) algorithm. Firstly, the paper introduces the bands used in cloud detection and analyzes the process of
feature selection. Secondly, special software named Libsvm was introduced, which was widely used in support vector
classification and support vector regression. Then the new method is established by building a model of remote sensing
image classification based on SVMs. The performance of SVMs was compared with the prevailing method of error back
propagation neural network (BP-NN) method with different training set numbers from 2000 to 250. The two methods
show similar detection accuracy when the training set number is larger (with a number larger than 1500), while SVMs
perform better than BP-NN method when the sampling number is smaller (with a number of 500 or less). There is no
significant difference of error rate among three kernel functions of SVMs algorithm. The general error rate is about 4%.
Thirdly, in order to valuate the capability of SVMs, two cases were selected for cloud detection using SVMs method. The
cloud was clearly identified either from land or sea, and the snow cover existed in both cases, which was selected
intentionally to test the capability of distinguishing the cloud from the underneath snow. Therefore, the SVMs technique
is proved effective as compared with traditional methods in remote sensing image classification and is worthwhile to be
popularized in the society of remote sensing applications.
Focused on the cloud detection task using EOS/MODIS information, this paper introduced a new method of cloud
detection by use of the Support Vector Machines (SVMs) algorithm. The performance of SVMs was compared with the
prevailing method of BP neural network (BP-NN) method with different training set numbers. The two methods show
similar detection accuracy when the training set number is larger (with a number larger than 1500), while SVMs perform
better than BP-NN method when the sampling number is small (with a number of 250 or less). SVMs method was then
used to detect cloud over both land and sea; it distinguished cloud from snow cover, water body, and other land surface
objectives clearly. Therefore, the SVMs technique is proved effective as compared with traditional methods in remote
sensing image classification and is worthwhile to be popularized in the society of remote sensing applications.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.