FAST RESIDENTIAL AREA EXTRACTION
REMOTE SENSING IMAGE BAS
ALGORITHM IN HIGH RESOLUTION
ED ON TEXTURE ANAYSIS
SU Junying * Qingwu HU ij
School of Remote Sensing and Information Engineering, Wuhan University, Wuhan, P.R. China, 430079
KEY WORD
ABSTRACT:
In this paper a residential area texture description based on the
based on the grey symbiotic matrix and Fourier
original residential area in the imag
area is very similar to the backg
processing of the texture characte
possess accordant grey value and
obtain through the histogram processing
interior texture change and asymmetric texture re
they can be changed from the asymmetric reg
image for the successful segmentation. And
experiment results of the semi-automatic e
resolution show this technique is very simpl
precision requirement of the mapping and su
r image such as the Gauss blur
or iterate processing
1. INTRODUCTION
High-resolution sensor technique has made great progress since
1990’s. The resolution of some Multi-spectrum remote sensing
images has reached 4 meters and that of some single spectrum
has been within 1 meter. High-resolution image can show the
object information such as Structure, texture and detail clearly,
which make it possible that observing object detail changes on
the earth through less space scale, large scale remote sensing
mapping and supervising the affection of human activity on our
environment. It's very important for the human residential area
planning and decision-making in economic region planning. A
residential area is a space entity composed of buildings, roads
between buildings, green areas, activity areas and production
areas, which has its definite structure, function and space shape.
Therefore, mastering the space distribution information of a
region's residential area will contribute to upholding
administration and application departments to make effective
decisions. For example, providing disaster estimation
department with space distribute information, helping to
understand human and region relation, serving for society,
economy and arts, upholding the supervision for living area
condition and the construction of living environment,
processing large scope semi-automatic or automatic mapping
with satellite image etc.
The researches of object automatic detection and recognition
(Zhang, 1996) are focused on the linearity object such as road,
river, electric power line and waters pound pipes for the
linearity object is the simpler object in the geometric objects.
The extraction of linearity objects can be realized through
image strengthen, slender and track processing. The extraction
technique of the difficult geometric shape object such as the
residential area has two Ways: automatic extraction and semi-
automatic extraction. Texture feature is the direct embodiment
of the object structure and space arrangement in the image.
Some researches are focused on the residential area extraction
of the satellite images while the precision is greatly depended
on the image quality and cannot reach the application
requirements. Yang, 2001 study to extract and recognize the
residential area with its spectrum character and space
* Corresponding author: Tel.:0086-27-87210286; Fax: :0086-27-876
S: High Resolution, Residential Area, Texture Analysis,
power spectr
€ is composed by the mixed pixels, which have inte
round, the texture character image is solved with the original image at first and a transform
presentation in the original imag
ion into equality region through the transform
a skeleton processing is proposed to eliminate
xtraction of the residential area in the remote
e and effective to the semi-automatic extraction o
rveying with satellite images.
Gauss Blur, Skeleton
3X3 region grey deviations is proposed which is much better than
um through the experiment and comparison data. Considering the
rlaced bright and dark pixels and the darkness
ation
character image
is applied to make the residential area in the texture
definite contrast relative to the background area so that the segmentation threshold can be easy to
in the texture character image. Some residential areas with large
€ have the same case in the deviation image while
ation processing to the texture character
the road from the residential area. The
sensing image with 3 meters ground
f the residential area and can meet the
distribution knowledge to TM images while the precision can
just reach 85 ?;. The research based on the nerve network (NN)
and the maximum similarity method can reach the precision
level of 69.1% and 523%. Computer based residential area
extraction includes two parts: recognition and measure.
Recognition can be easily realized for human but difficult to the
computer while in the contrast measure is easy for the computer.
The semi-extraction and recognition with human and computer
becomes the main study direction (Rob J.Dekker, 2001:
V.Bessettes, 1995).
In this paper a residential area texture description based on the
3X3 region grey deviations is proposed which is much better
than based on the grey symbiotic matrix and Fourier power
spectrum through the experiment and comparison data.
Considering the original residential area in the image is
composed by the mixed pixels, which have interlaced bright
and dark pixels and the darkness area is very similar to the
background, the texture character image is solved with the
original image at first and a transformation processing of the
texture character image such as the Gauss blur is applied to
make the residential area in the texture character image possess
accordant grey value and definite contrast relative to the
background area so that the segmentation threshold can be easy
to obtain through the histogram processing or iterate processing
in the texture character image. Some residential areas with large
interior texture change and asymmetric texture representation in
the original image have the same case in the deviation image
while they can be changed from the asymmetric region into
equality region through the transformation processing to the
texture character image for the successful segmentation. And a
skeleton processing is proposed to eliminate the road from the
residential area. The experiment results of the semi-automatic
extraction of the residential area in the remote sensing image
with 3 meters ground resolution show this technique is very
simple and effective to the semi-automatic extraction of the
residential area and can meet the precision requirement of the
mapping and surveying with satellite images.
64633; Email: Jysu_sjy@sina.com
1121
