The methods for supervised and unsupervised classification of remote sensing images are reviewed in this paper. The main problems discussed include the merits, shortages and application fields and conditions of these methods. An integrated evaluation of these methods is also given. The future developing trends and the key points in the study of remote sensing image classification are dealt with at the end of this paper.
There exist some shortages in the traditional image matching algorithm based on gray degree, such as huge quantities of calculation, relatively low accuracy, and too many restrictions in application. In order to solve these problems, this paper puts forward an optimized remote sensing image matching algorithm. The main ideas include the following several aspects: On the basis of confirming the subimage of the target image by understanding prior knowledge of remote sensing image, the first step is to search the subimage of the target image non-equidistantly with dynamic template, the second step is to locate the target position by two threshold gray degree correlation coefficients and conformal transform, and the last step is to judge the target position of the ground control point not recognized correctly by the spatial location relations of ground control points. The work flow is introduced in detail. Moreover, a comparison experiment on traditional and modified image matching algorithms is performed with an ASTER image and a TM image. From the results obtained, we can reach the conclusion that the modified algorithm is superior to the traditional algorithm in that it has much more higher accuracy and efficiency than the latter and hence it should have higher adaptability and applicability.
An algorithm for soil moisture change estimation is proposed by using S-band (3.0GHz) and VV polarization simulated backscattering-data based on IEM model. Such typical parameters as soil moisture, surface roughness and incidence angle are chosen to simulate two radar images for validation and improvement of the algorithm. The comparison results are well consistent with the inputting parameters under the applicable condition, which shows that the method can estimate spatial and temporal change of soil moisture.
Although optical remote sensing images have relatively high spatial resolution, they are frequently
disturbed by atmospheric conditions. If these disturbances can be eliminated or reduced, the application field of
these images will be extended. In this paper, some atmospheric correction methods have been described both in theory
and in application, and they are used on the same TM image. A comparison of the results has led the authors to
conclude that the algorithm presented by Liang Shunlin is better than other methods.
Leaf area index(LAI)is a crucial parameter of vegetation canopy structure and controls a number of
biophysical processes of vegetation. In this paper, a mixed model which combines the statistics model with the optical
model is presented to estimate LAI from Landsat-5 TM image data. Firstly, the model calculates and outputs a lookup
table (LUT) by useing of FCR model. and then, LAI mapping is conducted based on the empirical relations resulting from
the LUT. The results indicate that, being simple and easy to operate, the method can be used to estimate accurately
the LAI of reed marsh.
Based on the field survey data, this paper has analyzed the capability of Radarsat-SAR data on extracting
the forest stock volume. The results obtained show that the Radarsat-SAR backscatter coefficient has a relatively good
correlation with the forest stock volume, tree height and diameter at the breast height. Studies also show that the
backscatter coefficient is much more affected by height than by diameter at the breast height. It is held that C band
of SAR can be used to extract forest stem volume in a large plantation but is not suitable for forest survey.
This paper applied the Harmonic Analysis of Time Series (HANTS) algorithm based on Fourier transformation
to predigest and compress the NOAA/AVHRR Normalized Difference Vegetation Index (NDVI) time-series images, and at the
same time, wipe off and fill in the data where cloud cover and misdata are examined. The HANTS algorithm outputs such
Fourier components as amplitude components and phase components. A fit curve is built according to the Fourier
components. Temporal interpolation according to the curve makes it possible to reconstruct the sequential time-series
images.
For the selection of training samples in vegetation classification based on hyperspectral remote sensing, this paper investigates and analyzes the methods for selection of training samples in common use, presents two new selection (purification) methods for training samples, and then testifies the validity of these methods which are combined with specific OMIS-I hyperspectral remote sensing data.
This paper has classified the high-resolution SAR images obtained along the Huaihe River in July 2003. The authors first analyzed the high-resolution SAR image and pointed out the disturbing factors. After filtering the noise, the SAR image was decomposed with wavelet transform and the energy of the textures was computed by selecting a small window. Finally, the BP algorithm was used to classify the textures. The results indicate that, with the wavelet texture classification method, the high classification precision for the single-band, single-polarized and high-resolution SAR images can be obtained. This paper has also analyzed the shortages of the texture classification method for high-resolution SAR images and pointed out the research direction of the high-resolution SAR image classification.
Based on experiments and theoretical reasoning, the authors have analyzed quantitatively the relationship between the desertification and vegetation index (NDVI) and the land surface temperature (LST). The combined information contained in the LST-NDVI space is extracted to form a new index, namely difference index of desertification (DDI), which is then used to identify the severity of desertification. With this index, we can easily get to know biophysical properties of the land surface.
In this paper, some preliminary achievements on the application of remote sensing and GIS technology to regional agricultural and ecological studies are described. The WVS (Water, Vegetation and Soil)classification and its utilization in the land cover mapping at the regional scale are discussed. In addition, ecological evaluation has been carried out by using the regional geochemical data on the basis of the land cover map compiled from the Landsat ETM data.
This paper describes the investigation of inning and silting in Hangzhou bay using 8 times TM/ETM+ since 1986. Remote sensing image processes including geometric correction and false color combination and mosaiking were carried out before deriving shorelines. Then the information of shoreline dynamic changes, inning and silting was obtained by utilizing the space analysis function of geographical information system (GIS). The results show that the shorelines of Hangzhou bay have been changing since 1986 mainly due to manual inning and beach breeding. The investigation also reveals that the manual projects are carried out on the basis of planning.
Characterized by strong erosion of the south river bank from the Jingguang railway bridge in Zhengzhou to Dongbatou in Lankao, the lower reach of the Yellow River, which is a hanging river in this part, runs along Jiyuan—Kaifeng rift, with Taikang uplift on the southern side. With the discovery of Zhengzhou—Lankao regional concealed fault between Jiyuan—Kaifeng rift and Taikang uplift based on the interpretation of remote sensing and geophysical exploration information, the following phenomena can be explained reasonably: How does the river come into being, develop and evolve? What causes the river to stretch straightly from Zhengzhou to Lankao in the EW direction in spite of the fact that the river bed shifts in the river channel? Why does the river erode the south bank most strongly? The discovery of the fault and the river erosion characteristics is of great significance in river improvement and flood prevention in the lower reach of the Yellow River.
A new algorithm is proposed in this paper for fire detection with MODIS data. The crux of this algorithm is to combine brightness temperature of thermal infrared (TIR) bands with vegetation index derived from visible and near infrared (VNIR) bands for fire spot detection. Applications in China-Mongolia-Russia border area indicate that the algorithm is capable of detecting fire spots as well as reducing false alarms caused by non-vegetation areas.
On the basis of RS and GIS technology, Such indexes as mean area per patch, fractal dimension, patch
elongation, diversity, contagion, fragmentation and the TM images in 1997, 2000 and 2002 were used to study the
regional landscape characteristics and dynamic changes of the northern part of Chongqing city. The results show that
the regional landscape characteristics experienced deep changes in five years, from an urban-rural transitional zone
to a rapidly urbanized landscape. Various kinds of urban areas are expanding persistently and become spatially
centralized. On the other hand, the cultivated land and woodland were divided up and nibbled at, which damaged the
regional ecological environment.
The airborne remote sensing technique, combined with 3S technique and field survey, can play a major role
in dynamic monitoring for performing removal project during Three Gorges immigration. The steps of the removal project
were exactly reflected by the dynamic monitoring based on the remote sensing technique. This paper gives the dynamic
monitoring example of the second removal project in Badong, Wusan and Yunyang counties. The dynamic monitoring result
can offer timely and exact information for project supervision, and its data set up a foundation for dynamic
monitoring in future.
This paper has analyzed the application of three-dimensional visualized and animated picture simulation
technology to the dynamic simulation of the formation of geological scenery, with the emphasis placed on the
simulation of the formation and evolution of geological scenery at the temporal scale. Based on various modeling tools
in 3DS MAX, the authors fully used data and documents available to make dynamical simulation of the formation of
geological scenery, and attained lifelike effect. The applicability, potential and related techniques for the
application of 3DS MAX to dynamical simulation of the formation of geological scenery are also explained on the basis
of studying the dynamic simulation of the volcanic formation in the Changbai mountains. A satisfactory effect has been
achieved in this study.
The authors have applied the thought of image data base to creating the Large-Scale Three-Dimensional
Remote Sensing Image (L3DRSI), and put forward the new method “dividing-integrating”. The specially processed large
DEM data are divided into small blocks, and texture mapping of RS image with these DEM data is performed in two ways.
Finally, the L3DRSI is created. The new method can not only improve the precision of terrain which is very important
for L3DRSI, but also enrich the information and enhance the practicability of LRSI3D.