310
AN EXPERIMENT OF HIGH RESOLUTION SAR IMAGE
IN DYNAMIC MONITORING THE CHANGE OF CONSTRUCTION LAND
CaoYinxuan 3 ’* 5 , Zhang Yonghong b , YanQin b , ZhaoZheng b
a Key Laboratory of Land Use, Ministry of Land and Resource
b Chinese Academy of Surveying and Mapping
caoyx@casm. ac. cn
KEY WORDS: SAR, filtering, change information, dynamic monitoring, PCA, fusion
ABSTRACT:
This paper adopts high resolution SAR images and optical multispectral images of two dates to detect land use changes. Firstly,
different speckle filtering methods are applied on SAR images, and different speckle noise suppression effects are analyzed and
compared. It is indicated that the GAMMA speckle filtering method can get the best result. Secondly, in order to automatically
detect new-added construction land change information, some methods such as data fusion technology and false color composite
based on Principal Component Analysis (PCA) are put forward. Finally, change information is extracted.
1. INTRODUCTION
It’s the key technology of land resource and ecology
environment dynamic monitoring to acquire land use and land
cover change space time distribution information speedily and
correctly through advanced remote sensing image processing
technology based on different times images[l]. Recently, the
key objective of land use dynamic monitoring project is to
monitor the new-added construction land. Its aim is to acquire
the concrete position, boundary and attribution of new-added
construction land within monitoring area timely and efficiently,
and then provide beneficial technical support and guarantee.
Synthetic Aperture Radar (SAR) is independent of solar with
the characteristic of all weather, all time and penetrating cloud
and fog. Due to the geometry shape of building on SAR images,
its signal is more outstanding like the effects of comer
reflector 121 . Then it benefits the recognition of construction land.
And it makes SAR images be the more valuable data source of
dynamic monitoring during the special time 131 .
The application experiment on dynamic monitoring through
SAR images is already carried out in several years, but the data
source is mainly based on spacebome SAR data. And it is rarely
reported on high resolution airborne SAR data. This paper
introduces the experiment data sources based on airborne SAR
data in 2002 and 2004 and SPOT5 multispectral image in 2002.
And then new added construction land change information is
extracted through the combination between automatically and
artificial interpretation. The resolution of SAR data is about 2m,
and the band is Ku, and the resolution of multispectral image is
10m.
2. PROPRECESSING
2.1 Data correction
The test area locates in the conjunction between city and village.
It is even, and the topography fluctuation is small. So the
polynomial correction methods are put forward on SAR data.
The topography map is the reference base map as selecting
control points, and the mean square error is no more than 1
pixel. The result meets the correction accuracy requirement.
Then regarding the corrected SAR image as the reference data,
other image matches it correctly. Image registration is the base
of change information extraction. And if the error is too large, it
will bring more false change information, and reduce the
accuracy ratio of change monitoring.
2.2 Speckle filtering
Objects accept the coherence electronic waves emitted by radar,
and it includes many pulse scatter points, then electromagnetic
waves reflected are the coherence overlay of each scatter waves,
so it produces many speckle noises in the SAR images. The
existence of speckle noises reduces SAR image correction and
interpretation precision 141 . Therefore, in the practical application,
speckle filtering is the fundamental work of edge detection,
object classification, and target detection and recognition etc. 151 .
Recently, the research on speckle filtering is the hot topic of
microwave remote sensing. The aim of speckle filtering is to
reduce speckle noises and keep the texture detail information.
At the stage of speckle noises, there are different methods
according to different application purpose. The common
speckle filtering mainly includes: Gamma-MAP, Lee filtering,
SIGMA filtering, mean filtering, Frost filtering.
In the test, different filtering methods are carried out. Through
the contrast and analysis of filtering results on the speckle
reduction and detail preserve, the Gamma-MAP is finally
selected. Gamma-MAP filtering is built on noise model. The
following figure 1 is the contrast images between the original
image and filtering image, a is the original image, and b is the
filtering image through 3x3 window Gamma-MAP.
3. CONSTRUCTION LAND RECOGNITION ON SAR
IMAGE
In the SAR images, construction land reflection energy is high,
so it’s brighter than surrounding other objects. Otherwise, due
to residential building’s orderly generally arrangement, it
