The Hawaiian Islands contain more than two-thirds of the global life zones delineated by Holdridge1. We used high-fidelity imaging spectroscopy and shortwave-infrared (SWIR) spectral mixture analysis to analyze the lateral distribution of plant tissues and bare substrate across bioclimatic gradients and ecological life zones in Hawai'i. Unique quantities of photosynthetic and non-photosynthetic vegetation (PV, NPV) and bare substrate identified fundamental differences in ecosystem structure across life zones. There was a nearly 20-fold increase in PV fractional cover with a 10-fold increase in mean annual precipitation (< 250 to 2000 mm yr^-1). NPV fractional cover remained nearly constant at ~50% in ecosystems with a mean annual precipitation < 1500 mm yr^-1. Thereafter, NPV steadily declined to a minimum of ~ 20% at 3000 mm yr^-1 of rainfall. Bare substrate fractions were highest (~50%) at precipitation levels < 750 mm yr^-1, then declined to < 20% in the 750-1000 mm yr^-1 zones. The combination of low bare substrate and high NPV cover in the 750-1000 mm yr^-1 rainfall zones identified these areas as high fire risk. The results verify the applicability of SWIR imaging spectroscopy for ecosystem research on a global scale. They also set the framework for continued studies of ecosystem structure, function and invasive species throughout the Hawaiian Archipelago.

The Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) consists of three subsystems divided by the wavelength region: Visible and Near-Infrared (VNIR), Shortwave Infrared (SWIR), and Thermal Infrared (TIR) subsystems. The VNIR, SWIR and TIR subsystems have 3, 6, and 5 spectral bands with the spatial resolution of 15, 30, and 90m, respectively. The purpose of this study is to propose an algorithm for generating SWIR and TIR imagery with a 15m resolution based on spectral similarity. In the algorithm, SWIR images are first super-resolved using VNIR images, and TIR images are then super-resolved using VNIR and super-resolved SWIR images. The first step is as follows: 1) degrade the resolution of the VNIR images to 30m by pixel aggregation with the point spread function (PSF) of SWIR, 2) generate a homogeneous pixel map with a 30m resolution from the original VNIR images, 3) generate a multi-way tree for VNIR and SWIR spectra by stepwise clustering for the 30m-resolution VNIR and SWIR images, 4) generate super-resolved SWIR images by allocating the most likely SWIR spectrum to each 15m-resolution pixel based on spectral similarity in VNIR using the 30m-resolution VNIR and SWIR images, and the multi-way tree, and 5) modify the super-resolved SWIR images using the PSF as to be fully restorable to the original images. The second step is similar, except that super-resolved TIR images are derived from both the VNIR and the super-resolved SWIR images. In the latter part of the study, the algorithm is validated using ASTER data.

This paper addresses the issue on automated registration of images from weather satellites. Traditionally, weather satellite community has employed an approach called landmark detection for automated registration. A ground point or feature with known reference coordinates is defined as landmark. A landmark is matched against a weather satellite image. Based on match results estimated is the mapping function between images and a reference datum. This landmark detection approach has been suffered from the problem of mismatches. If match results contain errors, the accuracy of estimation deteriorates. To overcome this problem, we propose the use of a robust estimation technique called randam sample consensus (RANSAC). Through intelligent strategy this robust estimator will distinguish inliers from outliers and establish the mapping function with inliers only. This estimator has been reported to work in land observation satellite applications as well as in many computer vision applications. We will show that the RANSAC can also work for our purpose. We tested our approach using a global coastline database anda GOES-9 image. A global coastline database was processed to generate 30 landmarks. They were matched against a GOES-9 image. Visible inspection revealed that the results contained 13 mismatches. With 30 match reults the RANSAC was applied. It identified all 13 mismatches correctly. We can conclude that the RANSAC is able to select correct matches. For reliable automated registration, the RANSAC needs to be incorporated in the landmark detection process of weather satellite images.

In category classification of remotely sensed imagery, it is important that pixels of image are classified using spatial informaton. We have implemented MRF(Markov Random Field) model for a classification of higher accuracy. The model of MRF is a random field whose random variable is owed to its neighborhood. The LANDSAT TM data of the Kanto area, Japan, has been alalyzed with the manner of iteration in which probability density function for a confiuration of classes reaches a maximum. Partly because of taking into account of edge information in image, the results show considerably good classification.

In the region covered by variable amounts of vegetation, pixel spectra received by remotely-sensed sensor with given spatial resolution are a mixture of soil and vegetation spectra, so vegetation covering on soil influences the accuracy of soils surveying by remote sensing. Mixed pixel spectra are described as a linear combination of endmember signature matrix with appropriate abundance fractions correspond to it in a linear mixture model. According to the principle of this model, abundance fractions of endmembers in every pixel were calculated using unsupervised fully constrained least squares(UFCLS) algorithm. Then the signature of vegetation correspond to its abundance fraction was eliminated, and other endmember signatures covered by vegetation were restituted by scaling their abundance fractions to sum the original pixel total and recalculating the model. After above processing, de-vegetated reflectance images were produced for soils surveying. The accuracies of paddy soils classified using these characteristic images were better than that of using the raw images, but the accuracies of zonal soils were inferior to the latter. Compared to many other image processing methods, such as K-T transformation and ratio bands, the linear spectral unmixing to removing vegetation produced slightly better overall accuracy of soil classification, so it was useful for soils surveying by remote sensing.

Recent autonomy experiments conducted on Earth Observing 1 (EO-1) using the Autonomous Sciencecraft Experiment (ASE) flight software has been used to classify key features in hyperspectral images captured by EO-1. Furthermore, analysis is performed by this software onboard EO-1 and then used to modify the operational plan without interaction from the ground. This paper will outline the overall operations concept and provide some details and examples of the onboard science processing, science analysis, and replanning.

Most spatial data organizations need automated conflation technology. For the same geographic area, an organization usually has several sources of spatial data, with each source differing in terms of available spatial features, attributes, resolution, accuracy, and other qualities. Processing multiple sources of spatial data, along with their respective differences and advantages, is a huge and ever increasing problem. The maintenance of spatial data is very costly and time consuming. This situation will become intensified when more and more digital spatial data are offered by using Internet technologies.In this paper the conflation technology for integration of spatial data from different source on the Internet is introduced. Conflation attempts to match the spatial data of the source and destination coverage. If the coverage have different origins, it is likely that the shapes, and even the locations of the features, do not match exactly. Most conflation algorithms only match similar features that are very close to each other. But indeed, spatial object is defined by its location, shape, attributes and relationship to others. Therefore, before the matching itself, the semantic relations, topological relations and geometrical matching technologies have to be probed. The research work is performed on road network, which are captured in different data models on the Internet. The approach is based on matching criterions between the spatial data of different data models. At first, the semantic relations have been considered, and then different data models compare with topological relations and select the similar data sets. The geometrical matching has been done in the selected data sets and chooses the best reasonable one for the conflation result. It can get more quickly speed than other conflation approach based on statistical investigations before.

The aim of this study was to develop and test a simple and efficient data fusion and integration procedure for updating land-use base map in China with high spatial resolution (1-4m) IKONOS images. The data fusion and integration procedure was divided three stages: (1) using image merger method to generate hybrid 1m spatial resolution multi-spectral IKONOS images to enhance further image analysis. Compared with IHS (Intensity-hue-saturation), PCS (Principal component substitution), Brovey transform and Wavelets transform, SVR (Synthetic variable ratio) merging method was the best method and used based on the merging results. (2) Integration of spectral, textural feature and land-use type of former base map basing on rule-classifier to detect possible land-use change. (3) Under the GIS (Geographic information system) environments, the possible change areas were checked by visual interpretation to ascertain the change and land use type through screen overlay of images and land-use base map. The procedures were tested in Liangxiang town, the suburb of Beijing city using IKONOS images of the May 9th, 2001 in 2002. The IKONOS images can satisfied the updating land-use base map of China according the pilot results of the geometric position accuracy, types accuracy and areas accuracy of change regions.

Geometric and radiometric correction, image processing, information extraction and the integration of remote sensing, GIS and GPS in the specific approach for dynamic monitoring of land resources in mountainous areas are discussed. A synthesized method combining the image difference approach with comparison post classification is employed and a monitoring system based on remote sensing, GIS and GPS are set up. Different illumination conditions are key factors influencing the spectral features in mountainous areas, thus the comprehensive analysis of DEM and NDVI are employed to restrain the influence of terrain. Errors also commonly generate in the registration of different temporal images and much change information is usually lost when the mean-value smoothing template is employed in the image processing in mountainous areas. To reduce the information lost, a regional auto-adaptive smoothing template is employed. As a case study, according to the specific characteristics of mountainous areas, the TM images acquired from both 1994 and 1996 are processed for land change detection in Renhe District, Sichuan. Field experiments for radiometric correction are conducted in the areas of 25 Km2 in this district. The changed areas are precisely surveyed and validated after the fieldwork in which the database of detailed land survey is acquired. Combined with Geological Information System (GIS) technology and Global Position System (GPS), a 3S-based dynamic monitoring system of land resources change information in Renhe District is established, which helps the data renewal and daily management. Finally, the key factors influencing the accuracy of information extracting in mountainous areas are discussed.

Support vector machine (SVM) is a newly learning machine. In the paper, it applied the SVM method to research on remote sensing multi-spectral classification using Landsat TM data. It selected the typical low-hill area as study site, which was located on the southern of the Yangze River, China. The land cover types were divided into six categories, which were the waterbody, the construction land, the paddy field, the woodland, the teagarden, and the bare land, etc. The classification of the study site using the Kohonen networks method was also given. The classification results show that classification accuracy of the SVM method is better than that of the Kohonen Networks method. Especially it could discriminate the woodlands from the mountainous shadow. In conclusion, the SVM method could gain higher classification accuracy using smaller training sample in low-hill area. It could also solve the confusion problems among the ground objects.

In the paper, experiments and analysis of three pixel-based fusion methods had been discussed. The fusion methods include IHS, PCA and Brovey transform method. The fusion experiments were carried out in two circs, that is, between Landsat TM multi-spectral data and SPOT-4 Pan data, Landsat TM multi-spectral data and IRS-C Pan data. From the fusion results, the definition of all fusion images were improved greatly compared to the Landsat TM image. Especially the linear ground objects are much clear, such as the roads, the residents, the bridges, etc. According to the fusion between Landsat TM data and SPOT-4 Pan data, the Brovey fusion method was the best one. The PCA fusion method was better than the IHS fusion method. According to the fusion between Landsat TM data and IRS-C Pan data, the Brovey fusion method was also the best one. But the IHS fusion method was better than the PCA fusion method. Maximum likelihood method of classification was carried out on the fusion result, and classification accuracy of the classification results were evaluated. From the evaluation result, it can be concluded that classification accuracy of the Brovey fusion result is the highest between Landsat TM data and IRS-C Pan data. Classification accuracy of the IHS fusion result is the highest between Landsat TM data and SPOT-4 Pan data.

This paper describes work being done at Raytheon-Santa Barbara Remote Sensing (SBRS) in the area of entropy reduction of remote sensing data on the National Polar-Orbiting Operational Environmental Satellite System (NPOESS) Visible/Infrared Imager/Radiometer Suite (VIIRS) instrument. The VIIRS instrument will produce the largest amount of data on the NPOESS satellite platform, and thus has the greatest impact on data rate. The VIIRS instrument produces 22 bands of radiometric and imaging data, which must be transmitted to the spacecraft without loss of data integrity. VIIRS uses an implementation of the RICE algorithm, along with spectral subtraction and data trimming that are described in this paper, to provide lossless data compression. This paper will also describe a simulation that predicted the data reduction performance and the resulting sensor data rates when VIIRS observes the earth from orbit. This paper will also describe the VIIRS implementation of the Fault Tolerant 1394 data network that utilizes the 1394 ASIC chipset developed by the NPOESS Integrated Program Office (IPO) and Northrop Grumman Space and Technology (NGST). This high-speed network will facilitate the reliable transmission of large amounts of compressed and uncompressed science and telemetry data from the VIIRS instrument to the NPOESS spacecraft.

The data of Gunung Ledang region of Malaysia acquired through LANDSAT are considered to map certain hydrogeolocial features. To map these significant features, image-processing tools such as contrast enhancement, edge detection techniques are employed. The advantages of these techniques over the other methods are evaluated from the point of their validity in properly isolating features of hydrogeolocial interest are discussed. As these techniques take the advantage of spectral aspects of the images, these techniques have several limitations to meet the objectives. To discuss these limitations, a morphological transformation, which generally considers the structural aspects rather than spectral aspects from the image, are applied to provide comparisons between the results derived from spectral based and the structural based filtering techniques.

In remote sensing, one is often interested in not only ascertaining the presence of certain resources or objects of interest, but also in determining their locations. Ground registration involves locating the target in sensor coordinates and performing a series of coordinate transformations to convert this location to earth coordinates. One application of this would be in preparing a scaled map showing the precise locations of the resources/objects of interest. To improve ground registration accuracy one can combine multiple looks from a single sensor and/or looks from multiple sensors. One advantage in utilizing multiple sensors is that one can fuse the measurements in such a way as to exploit the best characteristics of each sensor. This paper is applied to vehicular mounted remote sensing and presents the benefits obtained when combining radar and IR as a means of determining ground coordinates of the objects of interest.