294
THE STUDY OF REMOTE SENSING IMAGE INFORMATION EXTRACTION
TECHNIQUES BASED ON KNOWLEDGE
Wenbo Wu, Jiaojiao Meng, Yuping Chen, Jing Chen
School of Geomatics, Liaoning Technical University, Fuxin Liaoning, China, mengjiao714@163.com
KEY WORDS: Knowledge, Remote Sensing Image, Information Extraction, Spectrum Knowledge, Texture Knowledge, Shape
Knowledge, Spatial Data Mining and Knowledge Discovery
ABSTRACT:
The study of remote sensing image information extraction techniques are still the important field of remote sensing techniques, at
present, remote sensing image information extraction based on knowledge is the most promising direction of development, This
paper concerns with remote sensing image information extraction techniques. At the first, the research analyzes and describes
detailed characteristics of remote sensing information on remote sensing image form the principles of visual interpretation; Then it
sums up remote sensing image information extraction techniques based on spectrum knowledge, texture knowledge, shape
knowledge and other assistant knowledge respectively, and it elaborates in depth on remote sensing image information extraction
techniques based on knowledge form several aspects, including the concept of knowledge, the discovery of knowledge and the
attainment of knowledge; Finally, it sums up the method of attaining knowledge automatically — Spatial Data Mining and
Knowledge Discovery. Conclusion is that remote sensing image information extraction based on knowledge is not only helpful in
synthesizing the spectrum and structure characteristics of image, but also in integrating the geographical principles and experience of
the distribution of objects on the earth, and it conduces to improving precision of remote sensing image information extraction.
1. INTRODUCTION
The development of Spaceflight Technology, Sensor
Technology, Control Technology, Electronic technology,
Computer Technology and Communication Technology has
impelled the development of remote sensing technology greatly
since 1980's. However, the level of remote sensing application
lagged behind the development of spatial remote sensing
technique, the outstanding representation is the remote sensing
data form satellite fails to be utilized adequatelyfl]. The
cognitive deficiency of remote sensing information and the lag
of remote sensing image information extraction’s level make
the rich connotative knowledge in the remote sensing data fail
to be mined and utilized adequately, and bring about the
enormous waste of remote sensing information resources and
the fall of their applied value. So the ability and the efficiency
of information extraction will become one of the problems of
the future remote sensing application.
The remote sensing image information extraction techniques
based on knowledge not only can synthesize the spectrum
characteristic and structure characteristic of the image, but also
can synthesize the geonomy law and expert experience about
the ground feature’s distributes, and offer the effective approach
for resolving remote sensing image information extraction and
improving the accuracy of information extraction [2 l Remote
sensing image information extraction techniques based on
knowledge emphasize that it discovers the useful latent
information from the image database, these information are
connotative and unware previously. So using the theory of
knowledge will be helpful to solve the contradiction between
the quick increase of the remote sensing data and the difficulty
of the data processing and understanding for people. This paper
analyses several kinds techniques of remote sensing image
information extraction based on knowledge, and discusses the
techniques of obtaining knowledge.
2. THE REMOTE SENSING IMAGE INFORMATION
EXTRACTION TECHNIQUES BASED ON
KNOWLEDGE
2.1 The Remote Sensing Image Information Extraction
Based on Spectrum Knowledge
What the remote sensing image takes notes is the ground
feature’s electromagnetic radiation information, and the
spectrum characteristic that are sounded and taken notes by the
remote sensing sensor is a basis, and is the intrinsic
characteristic of differentiating between two different kinds of
ground feature, and is the most important and the most direct
information in remote sensing image data. The spectrum
knowledge of the ground feature is the most important
knowledge in remote sensing information extraction, so the
study of the ground feature’s spectrum characteristic received
the each country’s recognition.
The spectrum characteristic includes spectral resolution and
radiation resolution The spectral resolution is the number,
wavelength and width of wave bands what remote sensing
sensor used. The radiation resolution is the intensity (the
brightness or the grayscale) value of the ground feature’s
reflected electromagnetic wave in a wave band. Generally, the
high spectrum resolution and radiation resolution can improve
the ability of distinguishing from ground feature effectively.
For the multispectral remote sensing images, carrying out
information extraction with the spectrum knowledge is very
effective. The spectrum knowledge includes the primitive
spectrum knowledge and the variational spectrum knowledge.
Using the primitive spectrum knowledge: First take samples for
various ground feature’s brightness value, compute the mean
value, and then discover the relation between the spectrum
which is able to distinguish form ground feature, build the
model of extracting such ground feature, carry out information
