Image processing techniques have been applied to Landsat TM images of Antarctica to enhance both the spectral contrast and the spatial information. The test site is located west of the Terra Nova Bay area, south of the terminal part of the Priestly Glacier and includes mountainous ranges characterized by alpine morphology, a large plateau and valley glaciers. Rock units are mainly constituted by a metamorphic complex, granitoids and supraglacial morains. A method based on principal component transform and on high-pass convolution filtering has allowed us to produce an image map where the spectral and the spatial information of the different surface units were combined. The detection of outcrop boundaries has been improved as well as the visual interpretation of their morphologic features. In general rock units correspond to those mapped by means of geological field survey. Morainic deposits are well discernable from the 'in situ' material appearing different, after processing, in texture and color from the intrusives and metamorphic complexes.

FUYO/OPS-shortwave infrared (SWIR) bands have been assessed on their ability to discriminate and identify minerals of the Al-OH and Mg-OH groups, and to separate distinct carbonates from each other as well as from non carbonatic rocks and soils. This research also focuses on the possibilities of discriminating and identifying mineral assemblages in hydrothermal alteration halos, including Fe^2+,3+-bands in the visible and near infrared range (VIS/NIR). Results are compared to LANDSAT TM data to gauge the examined technical parameters of OPS level 2 data and to investigate the effectiveness of three distinct subdivisional SWIR bands versus one in the 2.2 micrometer atmospheric window. FUYO/OPS data show comparable results to LANDSAT TM data for entropy, SNR, LSF and spectral corrrelation for the VIS/NIR bands. Although significant image defects, like blur and response delay problems, affect the three OPS-SWIR bands centered within the 2.2 micrometer window, a distinct separation of features producing minerals like kaolinite, alunite, gypsum and carbonates could be obtained from an arid test-site in the Negev desert using image processing and data extraction methods. The OPS-SWIR bands in combination with the VIS/NIR bands allow for an improved discrimination of alteration halos as compared to a single SWIR band (TM).

This paper discusses the application of stochastic labeling of remotely sensed images. A cooperative, iterative approach to segmentation and model parameter estimation is defined which is a stochastic variant of the expectation maximization (EM) algorithm, adapted to our model. Classical statistical modeling forces each pixel to be associated with exactly one class. This assumption may not be realistic, particularly in the case of satellite data. Our approach allows the possibility of mixed pixels. The labeling used in this technique involves two parts: a hard component, which describes pure pixels, and a soft component, which describes mixed pixels. The technique is illustrated by the classification of a SPOT HRV image. Because of the high resolution of these images, the pixel size is significantly smaller than the size of most of the different regions of interest, so adjacent pixels are likely to have similar labels. In our stochastic expectation maximization (SEM) method the idea that neighboring pixels are similar to one another is expressed by using Gibbs distribution for the priori distribution of regions (labels). This paper also presents a statistical model for the distribution of pixel values within each region. The initial parameters of the model can be estimated by using a K-means clustering or ISODATA, in the case of unsupervised segmentation. These parameters are then modified in each iteration of SEM. In the case of supervised segmentation, the initial parameters can be obtained from a classifier training data set and then re-estimated in SEM method. The reason for this re-estimation is that a set of classification parameters obtained from a classifier training data set may not produce satisfactory results on images which were not used to train the classifier. Our study shows that this SEM method provides reliable model parameter estimators as well as segmentation of the image.

Complex dielectric constants determined by inversion of the polarized returns of AIRSAR radar images acquired in wet conditions clearly delineate the distribution of saline soils in the Tragowel Plains irrigation area of Victoria, Australia. There is good agreement between the areas delineated as having anomalous dielectric constants by the radar inversion techniques with saline areas as defined by geophysics and determined in the field. The radar-determined complex dielectric constants are significantly smaller than might be expected from the known moisture contents of the soils at the time of image acquisition. The magnitudes of P band radar-determined dielectric constants most closely approach those expected from field determinations although the distribution of L band-determined dielectric constants give the best discrimination between saline and non-saline areas. Modification of the inversion algorithm to allow for the effects of vegetation produces complex dielectric constant magnitudes in the saline regions that are closer to expected values. Variations in the complex dielectric constant derived by inversion of SIR-C imagery acquired under dry conditions is largely a function of soil moisture content. This is closely related to the local recent irrigation history. Inversion of imagery acquired at small incidence angles does not yield meaningful results and it is inferred that this is because the small perturbation model does not apply. The results support the belief that incipient soil salinity can be mapped and monitored by polarimetric radar.

The aim of the research presented here is to point out the possibility of comparison and integration of the information derived from optical and SAR sensors for geotectonic studies. The study area, known as Fossa Bradanica, is located in Southern Italy and is a tectonic depression directed NW-SE, that represents the Apennine foredeep. It is characterized by a thick sequence consisting of clay, sand, and conglomerate strata. Optical and microwave images collected during three different seasons (spring, summer and winter) have been selected. Only the winter images have been interpreted because the winter Landsat-TM image proved to be more proper in tectonic studies, due to the low sun elevation enhancing the morphological characteristics of the area. According to this choice, winter ERS-1 images registered during both ascending and descending orbits have been analyzed. The identification and analysis of tectonic lineaments have been performed separately on the different images, manually detecting the lineaments by photointerpretation. The comparison of results points out that the synergistic use of the two sensors could offer an effective improvement in the knowledge of geological and structural data. 15

Usually the analysis of high resolution satellite images such as radar SAR ERS-1 images is undertaken by photo-interpretation techniques in order to reveal geological features. The numerical image processing is based on a filtering method designed for a better identification of geological structures on SAR images. The method leads to a mapping of recent faults on which the vertical offset is quantified. As examples, steeply dipping active faults with abrupt scarps are extracted from SAR-ERS1 images of the Central Andes (Atacama Fault zone, Northern Chile). The fault throws are then evaluated with a specific numerical image processing.

This paper presents an integrated use of cartography and remote sensing imagery supplied by satellite and aircraft to study the geomorphological aspects of an alluvial plain for archaeological purposes. The study area is located at the confluence of the Valtellina (Adda River) and Lower Mera River valleys in northern Italy. Landsat data and aerial photographs were used to study the partial filling of the Lake Como lacustrine basin resulting from the progradation of the Adda River delta. Different soil humidity content, related to variable grain size of the alluvial deposits is an indicator of ancient river beds which were formed in this area before the nineteenth century artificial rectification of the River Adda's final stretch. Profiles coincident with geological sections gained by geophysical sounding were performed on the remote sensing imagery to verify eventual correspondence of depositional features with different analysis techniques. The integration of remote sensing multilevel data with cartography and archaeological evidences has been useful for the assessment of the paleoenvironment which conditioned human settlements.

In order to advance the state-of-the-art in the collection of imaging spectroscopy, the U.S. Navy Space and Warfare Systems Command sponsored the development and fabrication of a new generation, well calibrated hyperspectral imaging spectrometer. Called the Hyperspectral Digital Imagery Collection Experiment (HYDICE), the sensor was built by Hughes Danbury Optical Systems, Danbury, Conn., delivered for integration into the Environmental Institute of Michigan's (ERIM) CV-580 aircraft in December 1994, tested and characterized between January and June 1995, and has since been involved in several airborne data collection experiments. In this paper, the HYDICE Program Office organization, sensor specifications, airborne characterization results, and a summary of the results of the most recent data exploitation and analyses are presented.

Remotely sensed data appears in databases in many forms and is used for a wide variety of purposes. Key questions arise in the use of such data bases and imagery such as 'who owns the data,' and 'what is the liability of those who provide the database for others to use.' In America, (where we litigate everything) these are important questions if you are to establish your rights to use remotely sensed data and, of most import, profit from that use. The law in the U.S. is anything but clear. There are however certain guidelines, that are discussed, that give you a better chance at avoiding problems and profiting from your technology.

Policy changes in the United States and Europe will bring a number of firms into the remote sensing market. More importantly, there will be a vast increase in the amount of data and potentially, the amount of information, that is available for academic, commercial and a variety of public uses. Presently many of the users of remote sensing data have some understanding of photogrammetric and remote sensing technologies. This is especially true of environmentalist users and academics. As the amount of remote sensing data increases, in order to broaden the user base, it will become increasingly important that the information user not be required to have a background in photogrammetry, remote sensing, or even in the basics of geographic information systems. The user must be able to articulate his requirements in view of existence of new sources of information. This paper provides the framework for expert systems to accomplish this interface. Specific examples of the capabilities which must be developed in order to maximize the utility of specific images and image archives are presented and discussed.

TRW is developing a World Wide Web based hyperspectral remote sensing data archive to provide diverse user communities with easy to use archive search and timely data transfer capabilities. This paper describes our initial version of this archive. Searches can be by geographic area, time interval, and instrument type requirements. Interactive graphical maps help users locate data without requiring specific geographical coordinates. Browse images aid preliminary data content and quality assessment. On-line data transfer is supported, with optional lossy or lossless compression. A specially developed lossy algorithm provides the compression and fidelity needed to support dialup access.

Multisensory remote sensing requires the simultaneous registration and real-time processing of the time-varying multi-sensor image data. The frame-based programming technique was developed to provide the appropriate multi-sensor data management. Any frame-based processing system supports the automatic data updating since the output of any sensor has been changed. The visual programming of data flows is naturally available through the usage of this approach. The appropriate set of the frame types is formed to design the most generic multisensory framework. Finally, the problem of real-time, multi-processor implementation of the frame-based software architecture is addressed.

An overview of a robust and implementable compression algorithm previously developed for multispectral imagery is given. This three-dimensional terrain-adaptive transform-based algorithm involves a one dimensional Karhunen-Loeve (KLT) transform followed by two- dimensional discrete cosine transform. The images are spectrally decorrelated via the KLT to produce the eigen images. The resulting spectrally-decorrelated eigen images are then compressed using the JPEG algorithm. The key feature of this approach is that it incorporates the best methods available to fully exploit the spectral and spatial correlation in the data. The novelty of this technique lies in its unique capability to adaptively vary the characteristics of the spectral decorrelation transformation based upon variations in the local terrain. In addition several relevant practical issues are addressed. These issues include: (1) Handling the panchromatic sharpening band, (2) tradeoff between spectral and spatial fidelity, (3) different grouping of bands for spectral decorrelation, (4) impact of band misalignment on performance, (5) impact of dead and saturated pixels on performance, and (6) impact of preprocessing on performance.

A new prediction technique using luminance compensation with adaptive block size for multispectral image compression is proposed. The luminance compensation which is a function approximation technique is updated block by block for efficient coding. Existing compression techniques do not exploit the property that the spatial features of neighboring spectral bands are almost identical. In other words, the image content of each band usually has almost the same spatial shape even when the pixel values are different. The approach proposed here is a novel scheme that takes advantage of the relationship of almost identical spatial feature between bands. It is a prediction scheme based on the luminance compensation between neighboring spectral bands.

A key step in the controlling of a newly acquired satellite image of a known area, is the reliable recognition and accurate image space coordinate measurement of previously selected control features on the ground. This is an ideal case for the application of model-based vision techniques, since considerable information is known in advance about both the scene and the imaging conditions. Ideally, this information includes the location, three-dimensional geometry and surface materials of fixed scene objects, approximate camera location and orientation data, and illumination conditions such as sun angle and atmospheric effects. This paper addresses the questions of which scene objects to select as control, what portions of the available information to use, and how best to apply it in practical situations. The control of both monoscopic images and stereo pairs is addressed. A detailed example is presented to illustrate the principles discussed.

The model supported positioning (MSP) project will provide more accurate horizontal and vertical absolute position information for aerial imagery, using image understanding (IU) technology and photogrammetry in conjunction with ground control and three-dimensional (3D) object models. The MSP architecture consists of three principal elements: (1) in-line positioning function (IPF); (2) registration adjustment function (RAF); and (3) control feature positioning function (CFPF). The IPF is a production function, automatically processing imagery, control and camera data, to provide improved positioning capability. The RAF is the application function, where imagery exploitation benefits from improved camera data, and through internal operations, generates data to support analysis and maintenance of ground truth positioning. The CFPF develops the initial control package that is used as the IPF and maintains control data over time as sites are repeatedly imaged. Our design allows for all three functions, though logically linked, to be physically displaced.

The new space missions, devoted to the study of the atmosphere, often rely on interferometric principles and to obtain data directly usable from the scientist it is necessary to perform plenty of heavy mathematical operations, whose on board implementation was unrealistic up to some years ago. This paper describes the algorithms selected to process the data coming from a Michelson-Morley interferometer working in the infrared spectrum and devoted to meteorological studies. Further a complete electrical design is presented together with results of a bread board. The obtained results confirm the design feasibility. The work was developed under Italian Space Agency (ASI) contract.

Passive radio wave remote sensing of a moving object by multiple individual sensors has been studied. A time-varying matched (TVM) filter bank has been suggested for the target reconstruction and detection. The TVM filter tuned to the target velocity compensates the target motion, maximizes SNR, and enhances the target image. In contrast, mis-tuned TVM filters yield blurred target images. In the multi-target case overlapped blurred image tails may interfere with the target recognition process, and cause target detection false alarm. In this work we study the statistics of the interference due to the randomly overlapped target image tails. Inspired by S. O. Rice's early work on shot noise statistics, assuming a uniform-Poisson distribution of target velocity-position, we present the probability density function (PDF), a generalized representation of Rice's integral, of the image tail interference. We evaluate this generalized Rice's integral. In an example we present the numerical results of the integral for the intermediate target density case.

A low-cost, portable, video-camera system built by University of Bristol for the UK-DRA, RARDE Fort Halstead, permits in-field acquisition of terrestrial hyper-spectral image sets. Each set is captured as a sequence of thirty-one images through a set of different interference filters which span the visible spectrum, at 10 nm intervals: effectively providing a spectrogram of 256 by 256 pixels. The system is customized from off-the-shelf components. A database of twenty-nine hyper-spectral images sets was acquired and analyzed as a sample of natural environment. We report the manifest information capacity with respect to spatial and optical frequency drawing implications for management of hyper-spectral data and visual processing.