Taking Shanghai region as the study areas,by using thermal infrared data of multi-sources polar orbit meteorological satellite,the research selected 264 preprocessed images to couple with the synchronous air temperature data gathered from 52 weather stations and then built two kinds of air temperature retrieval models changed with seasons and times.Results showed that the stable models for retrieving the air temperature could be constructed based on large numbers of samples from multi-sources remote sensed data,and furthermore the retrieval precision could be improved by using a field-measured data to modified the model.

As an important complementarity of aerial remote sensing,the UAVRSS(Unmanned Aviation Vehicle Remote Sensing System) has many advantages in specific field.Firstly,the paper  described 4 times aviation experimentations,then based on the experiment images and aviation data of UAV,all images were  mosaicked to a panorama image.And the mosaicked image was analyzed from many factors for evaluating the image quality.Finally,some improvements to the UAV image quality were discussed.

Airborne LIDAR is a new technology that can obtain high\|precision three-dimensional geographical data quickly.Filtering process for LIDAR points cloud data is to separate the non\|ground points from the ground points.Several important useful filtering algorithms for LIDAR points cloud data\|\|mathematical morphology based filtering algorithm,slope based filtering algorithm,TIN based filtering algorithm,pseudo scanning lines based filtering algorithm,etc.have been introduced,discussed and contrasted.their advantages and disadvantages,the improving advice to each filtering algorithms also have been presented in the paper.

Snow cover， snow depth/snow water equivalent， surface soil frozen/thaw state and soil moisture are the key variables in the three cycles including energy， water and carbon cycles. In order to better understand the remote sensing techniques of above parameters， this paper presents a comprehensive review of the progress in remote sensing of snow， soil frozen/thaw state and soil moisture， including the methods and theories of snow cover， snow depth / snow water equivalent， surface soil frozen/thaw and soil moisture remote sensing monitoring from visible， microwave techniques and the integration of multi-sources of remote sensing. The research progress of these parameters is summarized， and the prospects of these parameters are also discussed. The capability of snow， surface soil frozen/thaw state and soil moisture with remote sensing has been demonstrated to be improved greatly due to the retrieval algorithms development based on from single-sensor to multi-sensor combination， single-band to multi-band integration， especially on the virtual satellites constellation. Long time series data set of these surface parameters about 40~50 years were generated， then these products provide our better understanding on surface response to global climate change， and accelerating the application into the research of hydrology， climate and carbon cycles. This review will be helpful for the application of key parameters retrieval in water cycle with remote sensing.

To take Wuhu in Anhui province for example.Firstly,the author analysed the water body in Wuhu and the spectral characters of the earth's surface,then,summarized every class separability of the water information and the earth's surface in every band of ASTER Remote Sensing Image,after repeated experiments and analysis constructed the extracting model of water body which based on threshold of bands and relation of spectrum:B2>B3,B1+B6<127,B3+B4<54 and B3<24.In the end,the author compared and evaluated the conclusion of extracting model of water body with the conclusion of unsupervised classification,supervised classification and NDVI.The result of experiment indicated that this method is feasible and more simple,but higher demands were needed in the selected samples of the analysing process of spectral characters.

Evapotranspiation retrievals in conditions of rugged terrain and arid/semiarid sparse vegetation are always the hotspot in estimation of regional evapotranspiration from remote sensing images.Concerning with the complex characteristics of the loess hilly-gully region,such as undulating topography,un-unique land use/covers and less vegetated land surface,etc,land surface characteristic parameters and flux was retrieved,and the instantaneous vegetation transpiration and soil evaporation was estimated with the TSEB parallel model by using the Landsat TM data for an experimental study site located on the conjuncture area of Shaanxi and Gansu where the terrain surface is very undulating with less vegetation developed,heavy soil and water loss loess plateau.And then the calculated result was merged into daily evapotranspiration,also the spatial pattern of the daily evapotranspiration was analyzed.Actual evapotranspiration of the study area was calculated with the extra resistance method and Penman-Monteith equation to compare with the TSEB parallel model estimated,which suggested that the TSEB parallel model can be used for the accurate estimation of evapotranspiration in loess hilly-gully region.

The signal-to-noise ratio (SNR) is one of the most important indices which can be used to evaluate the data quality obtained by a remote sensor.To a great extent,the SNR of an image reflects the SNR of the remote sensor.Several typical methods to estimate the SNR of optical remote sensing imagery are summarized in this paper,and their merit and restrictions are presented.And this paper also performs the comparison and analysis between these methods based on their own principles,from six aspects including the automatic computation,the computing time,the stability,the applicability,the suitable sensor category,and the uniformity of estimating areas.In addition,the paper points out that the comparison and analysis between methods in various specific applications should be done in the future.The study will help to choose a reasonable SNR estimating method aiming at different remote sensors and different types of remote sensing images.

There are kinds of methods for ortho-rectification in application of remote sensing images，including Collinearity Equation Model，Strict Geometric Model based on Affine Transformation，Improved Polynomial Model，Rational Function Model，Method based on Neural Network，and so on.But there is lack of system comparison among these methods.On the basis of introducing the principle of the methods above，advatanges and drawbacks about these algorithms are summarized in this paper.Specific emphasis is the mathematical derivation and algorithm design of FM.Tikhonov method is taken to the progress of computation  of RFM.Two kinds of algorithm based on neural network was taken in application of ortho-rectification.To compare accuracy and effectiveness between the above methods，we make some experiments.The result shows that: on the condition of the same GCPs distribution，Rational Function Model that can reach sub pixel accuracy is the best of all from the viewpoint of precision and can be used in practice in spite of its relatively slower speed.

With complicated natural conditions, multiplicity of crop structure, small and dispersive distribution of parcel, the accuracy of images with moderate and lower resolution can't meet the acquisition of crop yield forecasting. With improvement of new sensors of high resolution, remote sensing imagery of high resolution can provide more abundant information such as texture, hue and so on. However, the current object-oriented classification approaches are not mature, which have too much thresholds to be set and more complicated and difficult to be used commonly. Therefore, combining QuickBird high spatial resolution satellite imagery with the field investigation data as mainly auxiliary information as well as using the pixel-oriented maximum likelihood method, crop planting area was obtained step by step, applying the principle of multi-scale information extraction,a test was set in Mianyang, Sichuan province.The result shows that the accuracy of crop classification is fairly exciting.

Ulan Buh Desert is one of Chinas major deserts.In recent years its rapid expansion has seriously affected the local ecological security.Desert vegetation is the most important ecological protection barrier in this region.Gaining the knowledge accurately of the distribution of vegetation is important.Calculated NDVI,and integrated principal component analysis combined with Gray Level Co\|occurrence Matrix texture analysis to analysis the ALOS image in the reserch area.Using NDVI and mean texture as the classification indices,the article determined the appropriate threshold range,and abstracted the vegetation information by using the decision tree method.The result shows that the decision tree method could use texture and other auxiliary information effectively,and achieve better classification results compared with traditional classification method.

New method for automatic geometric rectification based on image matching in remote sensing image is proposed,which can solve the problem of lower-precision and time-consumption in traditional manual geometric rectification.First,the pictures contained ground control points (GCPs) are gained in the remote sensing image which has been processed with basic geometric correction method.Secondly,the accurate matching between remote sensing image and GCPs pictures are found.Then,the outliers of GCPs are detected.Finally,the remote sensing image is precisely corrected by the GCPs.Also in this paper,a new multicriterion image matching algorithm is proposed to detect precise matching.The experimental results demonstrate the robustness and effectiveness of the proposed automatic geometric rectification method.

Earth observation system and technology realized the global real-time observation and played an increasingly important role in acquiring space-time information of global surface and deep earth which provides condition for environment monitoring and earth system research. Development trend of earth observation system was discussed based on three aspects as high resolution development direction of earth observation satellites and sensors, networking earth observation satellites as well as integrated and coordinative development of earth observing system. At the same time, technology scopes and requirements were analyzed from the aspect of country investment, capabilities of equipments development and their potential application value in other fields, cost of development and operational needs of the time and so on.

In July,2008,during the large\|scale synchronous remote sensing experiment of “WATER:Watershed Airborne Telemetry Experiment Research” in Zhangye district,Gansu province,we did the adjacency effect measurement synchronous with satellite at Lanzhou University grass site,with the device designed by ourselves.The adjacency effect correction coefficient was computed using backward least square algorithm.Then it was imported into the actual reflectance transformation equation to correct the adjacency effect of the synchronous ASTER image.The quality of the result image was improved.Moreover,the result image was analyzed compared with that the SHDOM equation.The result showed that,both the reflectance and normalized difference vegetation index of the result image corrected with the adjacency effect correction coefficient had greater differences,meanwhile the spatial autocorrelation between pixels decreased obviously.〖JP〗

The interaction between vegetation and climate is a complicated process.Many vegetation models were developed to understand the mechanism of interacting between vegetation and climate and assess the effects of climate change on vegetation.Furthermore,vegetation models have progressed from Static Vegetation Models (SVMs) to Dynamic Global Vegetation Models (DGVMs).DGVMs include the dynamic biogeochemical models and dynamic biogeophysical models,which mainly simulate the vegetation physiological processes,the vegetation dynamics,the vegetation phenophase and the nutrient cycling.The widely used DGVMs in the world include LPJ,IBIS,VECODE,TRIFFID and so on.The present research on DGVMs mainly focuses on four points:① the improvement of the accuracy of DGVMs; ② the comparison among different models; ③ the integrated model of DGVMs and Climate Models; ④ the research on the Carbon Cycle Data Assimilation System (CCDAS).

Applying remote sensing to estimate vegetation biomass has been paid more attention.When using the remote sensing information parameters and other parameters to construct models,researchers often need choose some parameters as independent variables which have obvious influence on biomass.Therefore,identifying the number of independent variables and choosing independent variables play an important role in the reliability and precision of estimation models.From single-variable models,multi\|variables models and nonparametric models,the article reviews on correlation analysis of independent variables in estimation models of vegetation biomass based on remote sensing,which offers some referenced value for choosing the independent variables in estimation models of vegetation biomass.

The global environment change has become one of the three global environmental issues with the greatest global impact.Lake is of great ecological significance in regional ecosystem study.Dynamic change as one of three main lake environmental problems of lakes has became the key point of the research area.Remote sensing techniques,as scientific tools of rapid investigation and monitoring,have been widely applied in the research of dynamic change in lakes.Main progress around of the world in areas of lake dynamic change research by means of remote sensing techniques are introduced in the aspects of basic principles and remote sensing data sources,remote sensing classification and dynamic change monitoring methods of lake information | especially,the advantages and disadvantages of remote sensing data source and dynamic analysis methods are analyzed deeply.In addition,some existing problems and the development trends of the recent lake dynamic change research are discussed.

Rapid urbanization has changed land surface properties and land-atmosphere interaction,which further may impact on the climate and environment.It is important for urban sustainability to timely and accurate monitor regional or global urbanization process.Currently,urban area in large scale could be derived from satellite images,and DMSP/OLS has already been widely used to detect global urbanization due to its time series.However,its saturation in urban area largely limited its further application.Therefore,an improved urban index (VANUI) was used to detect urbanization process in JING-JIN-JI area from 2001-2012 to examine its fitness in mega-city area,further provide reasonable information for urban planning and urban sustainability.Compared with the traditional night lights index,VANUI is indeed to improve the monitoring ability for urban center and reduce the influence of signal saturation on parameters derivation.The results showed that VANUI could catch the urbanization process that happened in JING-JIN-JI area during last decades.Spatial pattern of urban expansion from 2001-2012 indicated that the urban area of JING-JIN-JI was doubled from 2001-2012,and the fasted region about urbanization happened in Tangshan,Hebei Province.Linear trend was also detected in urban development by VANUI index from 2001-2012,and the most important increasing period showed in 2004-2007 and 2009-2012.

Taking a localized area in Kunlun Mountain as an example,the digital elevation model (DEM) in the highly mountainous area is effectively generated with ERS-1/2 tandem radar interferometry.The accuracy of the achieved DEM is assessed statistically by taking SRTM DEM and USGS GTOPO30 DEM as the benchmark.The main reason affecting the accuracy of interferometry with ERS tandem radar is discussed.The experimental results show that ERS tandem radar interferometry can be used to successfully extract DEMs for highly mountainous areas.It is therefore recommended that ERS tandem radar interferometry might be applied to generate or update DEMs for the western parts of China.
 

With the rapid development of remote sensing and information extraction technology,it is possible to automatically update the geographic information.The road is important manmade object and is component of geographic information data.Road information exacted from remotely sensed imagery to update the GIS database has become a hot research spot for many years.This paper summarizes the course of development on road extraction from imagery,and review main methods of road extraction,template matching，knowledge method,Snakes model,dynamic programming,ridge and valley exaction method,morphological method,Hough transform,multi-scale method,method based on segmentation and other techniques for road exaction as well.The characteristics of various methods and insufficient of road extraction from remote sensing imagery were analyzed.Finally,subsequent studies were prospected.

Remote sensing has profound implication in ecosystem investigation and research.However,traditional multispectral remote sensing data usually focused on the mass screening.It was difficult to be used to inverse the complicated properties and biochemical parameters of ecosystem for the limitation of spectral resolution.Hyperspectral remote sensing,which measure large numbers of narrow spectral bands,could greatly improve the accuracy and types of input parameters of ecosystem models.In present study,we systemically reviewed the implication of hyperspectral remote sensing in ecosystem processes and properties based on expounding the principle and characteristics of hyperspectral remote sensing,and proposed the potential implication of hyperspectral remote sensing in ecology.

Water was the crucial factor that kept wetlands stable and health,to fast and accurately extract wetlands water body using remote sensing imagery became an important wetlands investigation,research and protection method.Considering the middle spatial resolution and high spectrum resolution,high positioning accuracy and low price,TM remote sensing imagery would be one of the main data for wetlands investigation,research and protection.The research carried out an experiment of extracting water from inland limnetic wetlands using TM remote sensing imagery based on many arithmetic,and concluded that spectrum classification was the best among the area accuracy index,extraction accuracy index and visual effect index,the following was single band threshold and vegetation index method,however,multi-band relationship and water index method was the worst.The main negative factor that affect extraction result was that wetlands water couldn't be extracted absolutely,which was caused by the resolution of imagery and complex hydrology condition of wetlands.Therefore,pixel unmixing and multi-source remote sensing data fusion would be the important techniques of promoting water extraction accuracy.

Vegetation coverage is one of the objective indicators that reflecting the basic conditions of vegetation,which is concerned by agriculture and ecology.A most potential and rapid method for measuring the vegetation coverage is to extract it from the digital photo of the land surface.However,there was a lack of mature method to realize this procedure precisely.This paper introduces a method that is imitating NDVI,which is used to process the digital photos and to extract the vegetation coverage rapidly.The accuracy of the proposed method is validated by the supervised classification to calculate the vegetation coverage.The comparison of the results of these two methods indicates that the imitating NDVI results are as good as the supervised classification ones.The imitating NDVI method is more rapid and automatic than supervised classification.This method proposed is valuable to precision agriculture practice.

Synthetic aperture radar (SAR) has become one of the important tools to map the shallow water bottom topography.In this paper,progress on remote sensing of the shallow water bottom topography by SAR has been reviewed.Especially,SAR imaging mechanism and SAR imaging theory of the shallow water bottom topography and the shallow water bottom topography mapping by SAR are mainly introduced.Meanwhile,we discuss some issues in present of the study on remote sensing of the shallow water bottom topography by SAR and the research orientation in future.

Shadow detection in high spatial resolution remote image is very critical for target identification and information recovering.In this paper,an object-oriented method is used to detect shadow area.Firstly,the image is processed by segmentation to attain the image objects, then the knowledge is established according to the analysis of the spectral character of the shadow area in the image.Finally,the shadow area could be extracted by using the rules defined in the knowledgebase.The experiment result shows that this method is valid.

For meeting the demand for large-scale agricultural monitoring system with remote sensing technology,extracting paddy rice information on the remote sensing image must be rapidly,precisely and reliable conducted.In this paper,paddy rice identification with SPOT-5 image was taken as an example on the 852 farm in Heilongjiang province of China.Firstly,the spectral characteristics of paddy rice and other six land-use types in this area were analyzed to find the possibility of extracting of paddy rice from the background.The results show it is difficult to distinguish paddy rice information from background on the SPOT-5 images because of complexity of spectrum and lack of band information.Secondly,taking those into account,characteristic bands for paddy rice extraction were proposed and merged into SPOT-5 images in order to increase spectral information and improve the separability.Thirdly,a simple model of decision tree was applied to extract paddy rice information.Finally,the results were checked by visual and statistica1 accuracy assessment.The results suggest that the model based on characteristic bands is simple and effective,and the accuracy by the model is much higher than that by the supervised classification method.However，some pixels in the neighborhood area between paddy rice and corn were misjudged.

This paper analyzes the key factors that result in the estimation errors of the MODIS land surface temperature products.Then the principle methods of validating the MODIS LST was discussed and compared.Normally there are bigger uncertainties of MODIS LST products in the semi\|arid and arid regions.As a case study,the Heihe river basin was selected to estimate the accuracy of MODIS LST products in Chinese arid and semi-arid regions.The two kinds of ground-measured observation data were obtained from the automatic meteorological stations:Infrared Radiation Thermometer data and longwave radiation data.The comparisons were performed for the validation based on these two types of observation data over three typical land cover types.The result demonstrates that:it is more reasonable to use long\|term nighttime surface longwave radiation data to validate the MODIS LST products.The validation results indicate that mean absolute error is less than 2.2 ℃ in the selected stations.

Night\|light data from the Defense Meteorological Satellite Program (DMSP) Operational Linescan System (OLS) has been widely utilized to derive the urban built\|up areas.Four major methods,including Empirical Thresholding,Sudden Change Detection,Statistics\|assisted Thresholding,and TM\|assisted Thresholding,have been proposed to determine the threshold value for separating the illuminated urban areas from the dark background of rural areas in DMSP/OLS night\|light images.This paper makes a comprehensive assessment of those methods through a case study of Shanghai,China.The methods are implemented to extract the urban built\|up area for Shanghai using DMSP/OLS stable light data acquired in 2003.Then,the same threshold values obtained from 2003 DMSP/OLS night\|light data are applied respectively to the 2000 and 2006 DMSP/OLS data,resulting in a significant error in urban built\|up area detection.This analysis result suggests that the threshold value determined for a specific year cannot be extended and transferred to other years.The failure in temporal extensibility of threshold value means that an appropriate threshold value has to be determined for every year when a time series of DMSP/OLS nigh\|light data need to be processed.Therefore,the method that determines threshold value independent of reference data is more suitable for processing time series DMSP/OLS data.The Sudden Change Detection method does not require ancillary reference data and is the best choice of those methods considering the convenience,accuracy,and automated data processing,and it is then adopted to derive the urban built\|up areas of Shanghai from 2000 to 2006.

Earth observing technology was used in civilian fields since the initial part of 1970s.Satellites ground stations are the basic establishments of earth observation system and basic source of space data,and they play the role of public data flat in national economic construction activities,and their development got wide attentions of governments.There 21 ground stations in different countries were investigated in this paper,including America,the Europe,Japan,Canada and Australia,etc.The properties,distribution characteristics,management system,personnel size and data types received of overseas satellites ground stations were analyzed in this paper.The results showed these overseas satellites ground stations had automatization,commercial,marke-oriented,civilian,internationalization,systematization,net-oriented operation characteristics.Future development direction of satellites ground stations was put forward in this paper.

The risk level of forest fire depends not only on weather,topography,human activities,socio-economic conditions,but also closely related to the types,growth,moisture content,and quantity of forest fuel on the ground.How to timely acquire information on the dynamics of growth and moisture content of forest fuel and climate in the whole country is critical to national-level forest fire risk forecasting.The development and application of Remote Sensing (RS),geographic information system (GIS),databases,Internet,and other modern information technologies has provided- an important technical means for macro-regional forest fire risk forecasting.Quantified forecasting of national-level forest fire risk was studied using Fuel State Index (FSI) and Background Composite Index (BCI).The FSI was estimated using MODerate resolution Imaging Spectroradiaometer (MODIS) data.National meteorological data and other basic data on distribution of fuel types and forest fire risk rating were standardized in ArcGIS platform to calculate BCI.The FSI and the BCI were used to calculate the Forest Fire Danger Index (FFDI),which is regarded as a quantitative indicator for national forest fire risk forecasting and forest fire risk rating,shifting from qualitative description to quantitative estimation.The major forest fires occurred in recent years was taken as examples to validate the above method,and results indicated that the method can be used for quantitative forecasting of national-level forest fire risks.

Three-dimensional （3D） radiative transfer model can accurately describe the interactions between solar radiation and heterogeneous land surfaces. Recently， it has become an important tool for quantitative remote sensing studies. LESS is a ray-tracing based 3D radiative transfer model， which take full advantage of the forward ray-tracing techniques for simulating radiative budget and backward ray-tracing for simulating large-scale images， which makes it possible to simulate various remote sensing data in a single model. Currently， LESS can simulate multi-angle Bidirectional Reflectance Factor （BRF）， multi-spectral/high-spectral images， fish-eye cameras， upwelling/downwelling shortwave radiation in rugged terrains and layered FPAR， etc. This simulated dataset can be used for validating physical modes， developing parameterized models， as well as training neural networks. This paper presents the fundamentals of LESS and its applications. LESS can be downloaded from www.lessrt.org.

To generate 3D information from the image sequence of the normal UAV(Unmanned Aerial Vehicle) loading camera,a reconstruction method is proposed.At the first step,the key frame selecting from the sequence,the robust features extracting and corresponding of multi-images of sequence are accomplished.At the second step,two views which are far away enough and sharing lots of correspondences are selected.With the correspondences the RANSAC algorithm estimates the fundamental matrix robust and then the more accurate correspondences with less outliers are gotten,which are called as inliers.Basing on the calibrated camera parameters,the motions R and T of the view is computed by decomposition of the fundamental matrix,and then it is optimized with the correct correspondences.At last,3D reconstruction is realized by the method of triangle communicating.Experiment results show that the algorithm is efficient and it improves the robustness and accuracy of the reconstruction.

In remote sensing image processing,edge extraction is our focus and difficulty.The paper showed the conception and species of the edge,and deeply explored the differential operator methods of first order and differential operator methods of second order.We contrasting and analyzing the principles and processes of every differential algorithm and the conclusion of the image processing.In the end,we bring forward our viewpoint about the problems the remote sensing image edge extraction technology is facing and where is the developmental direction of it.

Interference imaging spectrometry is in the frontier and hotspot of remote sensing technology and application.It has enormous application potential for its characteristics such as high luminous flux,high spectral resolution,and high-stability of imaging.We introduce the basic principles of interference imaging spectrometer.We will also give an overview of data processing technologies including spectrum reconstruction,calibration and atmospheric correction.Some representative spectrometers and their applications are reviewed in brief.Finally,it is pointed out that the effective approach to promote the development of interference imaging spectrometry shall be carrying out some research and applications based on HSI instrument loaded on HJ-1A satellite in our country.

Taking topographic map in 1970s as reference,four periods terrestrial coastlines in Zhejiang province were depicted by using TM data in 1986 and 1995,ASTER data in 2005 and HJ data in 2009.In those four years,changed coastline,increased land area and fractal dimension value were calculated.The conclusions were as follows: ① The three satellite data could meet the precise and quick requirements of rapid coastline change monitoring;② Zhejiang terrestrial coastline fast expanded forward ocean,205.24 km² newly increased land from 1986 to 1995,319.85 km² from 1995 to 2005,484.75 km² from 2005 to 2009;③ The  fractal dimension of coastline in Zhejiang province was relative small,which meant low level of complex,the value of fractal dimension was gradually decreased in recent two decades.
 

In recent years， in order to improve the classification accuracy of ground objects， break through the technical system of single sensor， and make up for the limitations of single data source application， multi-source remote sensing data fusion has become a research hotspot concerned by many scholars in the field of remote sensing. The fusion technology of optical image and LiDAR point cloud data of hyperspectral remote sensing technology provides a feasible scheme to improve the accuracy of ground object recognition and classification at the technical level， breaks the technical upper limit of single sensor， and provides a new solution for the integrated acquisition of target three-dimensional space-spectral information. At the same time， it lays a foundation for the research of hyperspectral LiDAR imaging technology. This paper reviews the development history of LiDAR and hyperspectral imaging data fusion， discusses the main fusion methods and research progress at the feature level and decision level， introduces the commonly used feature level fusion and decision level fusion methods in detail， summarizes the latest research algorithms and discusses their challenges and future development and application prospects. Finally， the future development of LiDAR and hyperspectral imaging data fusion is prospected systematically.

In the methods of improving the classification precision of remote sensing images,adding textural information as an expanded eigenvector into feature space is a pretty useful method.In this paper,the author extract texture using spatial connections between geo-objects,then put it into the classification process.This experiment shows a nice result.Through the problems encountered in this experiment,we discussed the appropriate scope of this methodology.

FY-3 MWHS scans the Earth's surface by mechanically rotating the antenna around X-axis.This research presents a geolocation method for FY-3 MWHS' remote sensing image based its scan geometry.It creates the spatial relationship model between the sensed data and the Earth based coordinate system,according to MWHS' sensing geometry and the sensor's attitude and position.This method is also including an algorithm of satellite orbit computation,which is used to get the satellite's instantaneous velocity vector from its position vectors.This method has been applied to FY-3 MWHS' remote sensing image geolocation.The geolocation result and error analysis use the land/water mask.Compared the geolocation results with the true remote sensing image,it shows that this method for MWHS' remote sensing image achieves pixel geolocation accuracy.

The high resolution sea surface wind speeds retrieval from SAR image is of great significance in the field of microwave remote sensing.A preliminary study about the high resolution sea surface wind speeds retrieval from SAR image combing with the NCEP/QSCAT blended sea surface wind directions was introduced.As a case study,this method was applied on one scene ENVISat ASAR imaging on Dec.6,2005.The retrieved wind speeds are compared with those from both the NCEP/QSCAT blended sea surface wind speeds and daily averaged Quick Scatter meter sea surface wind speeds.The results gave that they are in good agreement.The root\|mean\|square errors of wind speed are 1.9 m/s and 1.6 m/s respectively,which show that SAR has the ability of retrieving the high resolution sea surface wind speeds,and give the orientation of SAR sea surface wind field business application in future.

Satellite image fusion is always a focus in remote sensing field.According to data sources difference,image fusion can be divided into two broad categories:fusion of images using different sensors data and using same sensor data.Five recently proposed/modified fusion algorithms have been employed to test the fusion results of the two categories.The image pair,TM+ SPOT pan,was used to test fusion between different image sources,while the pair,IKONOS MS+ pan,was employed to test fusion between same sensor data.The study reveals that the overall results of fusion between same sensor data are better than those of fusion between different sensors image sources.The selected fusion algorithms have different performance in image fusion results between the two categories.The SVR transform is suitable for both categories of image fusion.It can greatly improve the spatial resolution,information quantity and clarity,but retain spectral information of the original multispectral image.The SFIM\|fused image has the highest spectral fidelity in both categories of image fusion,but has the lowest spatial frequency information gain.Although the MB transform can improve the spatial resolution of the original image,it generally failed to improve spectral fidelity,information quantity and clarity.The Ehlers transform is more suitable for image fusion between different sensors data while the WT is more applicable to image fusion between same sensor data.

With the development of convolutional neural network technology， recent research has paid more attention to the improvement of accuracy and the improvement of semantic information. Mask R-CNN network is a further improved segmentation network of Faster R-CNN. It has a good segmentation effect in high-resolution remote sensing image feature recognition. However， since the convolutional neural network can only be trained and predicted with small tile images， there is a large semantic information error in the prediction results. Faced with this problem， this paper proposed a gap-repairing algorithm based on the defect of prediction result of convolutional neural network. The approach use overlapsize algorithm to improve the matching degree between the prediction result and the ground-truth result at first. Then fill the gap through the correlation function in the PostGIS database to repair the small tile， which can make it be spliced ??into a complete picture. The research and experiment results showed that the algorithm could improve the image semantic information well and has practicability.