January 10, 2011 / Vol. 9, No. 1 / CHINESE OPTICS LETTERS 011003-1
Combined multi-kernel support vector machine and wavelet
analysis for hyperspectral remote sensing image
classification
Kun Tan (
nnn
) and Peijun Du (
ÚÚÚ
)
∗
Key Laboratory for Land Environment and Disaster Monitoring of State Bureau of Surveying and Mapping of China,
China University of Mining and Technology, Xuzhou 221116, China
∗
Corresponding author: dupjrs@cumt.edu.cn
Received October 25, 2010; accepted November 30, 2010; posted online January 1, 2011
Many remote sensing image classifiers are limited in their ability to combine spectral features with spa-
tial features. Multi-kernel classifiers, however, are capable of integrating spectral features with spatial or
structural features using multiple kernels and summing t hem for final outputs. Using a support vector
machine (SVM) as classifier, different multi-kernel classifiers are constructed and tested using 64-band
Operational Modular Imaging Spectrometer II hyperspectral image of Changping Area, Beijing City. R e-
sults show that by integrating spectral and wavelet texture information, multi-kernel SV M classifiers can
obtain more accurate classification results than sole-kernel SVM classifiers and cross-information SVM
kernel classifiers. Moreover, when the multi-kernel SVM classifier is used, the combination of the first four
principal comp onents from principal component analysis and wavelet texture provides the highest accuracy
(97.06%). Multi-kernel SVM is therefore an effective approach to improve the accuracy of hyperspectral
image classification and to expand possibilities for remote sensing image interpretation and application.
OCIS codes: 100.4145, 100.5010, 100.7410.
doi: 10.3788/COL201109.011003.
Hyperspectral remote sensing image classification re-
mains a contentious topic in remote sensing
[1−3]
. With
the increasing amounts of data from airbo rne and satel-
lite hyperspectral sensors, numerous bands and low
training samples pose a “curse of dimensionality”
[1−3]
.
To address this issue and to increase the stability of
classifiers, some feature selection algorithms have been
proposed. However, the proposed approaches, time-
consuming and scenar io-dependent, usually lead to infor-
mation loss
[4]
. In recent years, suppor t vector machines
(SVMs) have been widely used to classify hype rspectral
images, demonstrating excellent performance in terms of
accuracy, generalization, and robustness
[5]
.
According to related studies, SVMs have better perfor-
mance compared with traditional classifiers
[1,6,7]
. This
is observed when spectral information is use d for a SVM
classifier. Moreove r, SVMs can ac c ount for both spec-
tral information and spatia l features; for example, a
good framework of multi-kernels for hyperspectra l im-
age classification has been presented
[5,8,9]
. However, the
researchers did not pay mo re attention to the parameter
selection technique and more sophistica ted texture tech-
niques — two very important methods for improving the
performance of multi-kernel approaches.
Many recent studies attempted to enhance texture
analysis through different methods, such as texture clas -
sification
[10]
, texture segmentation
[11,12]
, and texture
detection
[11,13]
. Various texture ana lysis techniques have
also been developed. Traditional statistical approaches
to texture analysis, such as co-occurrence matr ic e s, sec-
ond or der statistics, and Gauss-Markov random fields
[14]
,
are restricted to the analysis of spatial interactions over
relatively small neighborhoods on a single scale.
The wavelet theory has developed rapidly since the
1980s, and has been used in fields such as sig-
nal processing
[11,15−17]
, image restoration
[18]
, image
retrieval
[19]
, and pattern recognition
[20]
, a mong others.
Uses of wavelet theor y on the texture analysis of re mote
sensing images have also been investigated
[21]
.
In this letter, SVMs, which have demonstrated supe-
rior performance in the context of hyperspectra l image
classification, are adopted a s the classifier. A series of
multi-kernel classifiers ba sed on SVM are constructed to
account for both spectral and spatial features (describe d
by wavelet texture information).
The SVM theory for a two-class problem is found in
some references
[22,23]
. Some p opular kernel functions in-
clude linear kernel, polynomial kernel, Gauss ian radial
basis function (RBF) kernel, and sigmoid kernel.
Considering some common SVM kernels, the bottle-
neck is defined as the kernel mapping function that at-
tains similar samples. In general, we can reconstruct a
new kernel if it fulfils Mercer’s condition.
Let χ be any input space and K : χ ×χ → R be a sym-
metric function. In the expression, K is a Mer c er’s kernel
if, and only if, the kernel matrix fo rmed by restricting K
to any finite subset of χ is positive semi-definite, having
no negative eigen values.
We can then consider the convex combination of mul-
tiple kernels:
K(x
i
, x
j
) =
K
X
k=1
β
k
K
k
(x
i
, x
j
), (1)
where 1 > β
k
≥ 0 and
P
k
β
k
= 1, and each kernel sat-
isfies the condition of Mercer’s kernel.
For definite pixel entity x
i
, if x
i
is recorded with x
s
i
in
the spectral doma in and x
t
i
is in the spatial domain after
1671-7694/2011/011003(4)
c
2011 Chinese Optics Letters