Estimation of foliar traits through imaging spectroscopy is one of the current major challenge in remote sensing to monitor ecosystem functions. New spaceborne hyperspectral sensors are upcoming or already in activity. In this context, future estimation models are required to be robust to temporal variations (e.g. illumination and phenological conditions) and changes in vegetation geometry in order to cope with the wide variety of canopy observations.Our work is threefold. First, we analyze the direct radiative transfer model to highlight potentials of one-dimensional CNN (1D-CNN) to retrieve foliar traits. Secondly, we build a 1D-CNN inverse model and search for an optimal architecture. Finally, we test the retrieval performance of our 1D-CNN on four oak species over four forest sites in California at different phenological stages during two years.

In preparation of the micro-bolometer-based MIcro Satellite for Thermal Infrared GRound surface Imaging (MISTIGRI) mission, we study the error budget of the Temperature-Emissivity Separation (TES) method using several spectral configurations that differ in channel numbers, locations, and widths. The error budget quantifies the contribution of 1) the TES underlying assumption about emissivity spectral contrast, 2) the errors on atmospheric corrections, and 3) the instrumental noise. When dealing with atmospheric corrections, we consider errors in atmospheric temperature, water vapor content, and concentrations of CO2and O3. To that end, we design an end-to-end simulator of MISTIGRI measurements in order to simulate the radiative and biophysical quantities involved in the data processing. We conduct numerous simulations over a wide range of realistic setups that include cavity effect, i.e., radiance trapping within vegetation canopy. In the case of micro-bolometer-based sensing, the current study highlights that atmospheric and instrumental noises have similar impacts on the TES retrievals, with resulting errors twice as large as those due to the TES intrinsic assumption about spectral contrast, where the latter contributes to the TES error budget within the [0.005–0.009] interval for emissivity, and within the [0.3–0.4 K] interval for land surface temperature (LST). Also, we show that retrieval performance of surface temperature is very similar across all considered MISTIGRI spectral configurations, with RMSE variation within 0.2 K. Eventually, our study permits us to select a 4-channels spectral configuration as the most suited for the MISTIGRI instrument, notably because it enables a moderately better capture of the emissivity contrast than a 3-channels one.

Hyperspectral pansharpening methods, which aim to combine hyperspectral and panchromatic images, yield limited performance for scenes whose strong spatial heterogeneity induces mixed pixels. The SOSU method has been designed to handle this limitation and provided good results on agricultural and peri-urban landscapes. However, its performance was reduced on more complex urban scenes, which contain a higher proportion of mixed pixels. This article presents a new version of this method, called SOSU-2021, adapted to better process urban scenes. SOSU-2021 is tested on an urban dataset at a 1.6 m spatial resolution. We obtain better numerical results than with the previous SOSU version, and in the worst case, 56 % of the mixed pixels are better or equally processed by SOSU-2021 than by the method used as a reference.

Forest characterization at tree scale is possible with remotely sensed images of high spatial resolution. A first step is the Individual Tree Crown (ITC) delineation which corresponds to the segmentation of canopy cover into different tree crowns. In this study, a new method based on an initial watershed segmentation of Canopy Height Model (CHM) is proposed. This method is based on a graph transformation of the initial segmentation map in order to apply structural criteria (calculated owning to CHM) and spectral criterion (computed on red, green and blue bands). Adaptive thresholds are defined according to the studied forest characteristics in order to improve low quality segmentation cases. This method has been applied on a complex forest site (Broadleaf dominant, steep relief…). The results show a 17% increase of global performance in comparison to the reference watershed segmentation.

Classification of hyperspectral scenes has been dominated in recent years by Convolutional Neural Networks (CNNs). Spectral and spatial convolutions have proven to be very effective in learning discriminative representations. The power of CNNs however is limited by their small receptive fields. Training deeper CNNs with wider receptive fields is a very difficult task with regard to the small amount of available training samples. Furthermore, CNNs seem to fail in capturing very local spectral features such as absorption peaks. In the present paper, we introduce a new paradigm inspired by physics-based hyperspectral indices and attention mechanisms. Our model learns spectral indices by focusing on specific spectral bands through soft attention units. It achieves high or better overall accuracy and kappa score than state-of-the-art CNNs on Pavia University, Kennedy Space Center and our own real-world dataset while dramatically reducing the number of parameters needed and increasing interpretability.

The atmospheric correction of remote sensing data in the reflective domain is today very well controlled under clear sky conditions. However, cirrus clouds represent 2/3 of the global terrestrial cover, making several images unusable. Gao and Li proposed an empirical method of thin cirrus correction. The method shows very good results on dark surfaces but presents bias higher than 0.02 when the surface becomes too reflective and the cirrus too thick. In addition, it only corrects for the upwelling path. Simulations show that the presence of cirrus on the sun-to-ground path has a significant influence on the received signal and must also be corrected. This paper shows that considering the transmission term of the cirrus of the upwelling and downwelling paths improves the results in the red-edge but also the SWIR, with an RMSE divided by 2.

In a very recent paper, we introduced (i) a specific hyper-spectral mixing model for the sea bottom, based on a detailed physical analysis which includes the adjacency effect, and (ii) an associated unmixing method, which is not blind in the sense that it requires a prior estimation of various parameters of that mixing model. We here proceed much further, by first analytically showing that this model can be seen as a specific member of the general class of mixing models involving spectral variability. Therefore, we then process such data with the IP-NMF and UP-NMF blind unmixing methods that we recently proposed in other works to handle spectral variability. Such a variability especially occurs when sea depth significantly varies over the considered scene, and we show that IP-NMF and UP-NMF then yield significantly better pure spectra estimation than a classical method from the literature which was not designed to handle such a variability.

Land surface emissivity is required for deriving surface temperature from thermal infrared radiances. When using single-channel or two-channel thermal infrared sensors, information on emissivity may be derived from spectral reflectance measurements through regression models. In this study, we present relationships derived over bare soils for Landsat 7 - ETM+ sensor. Reflectances in ETM+ channels were obtained from soil spectra (between 0.4 and 13 μm) extracted from the ASTER spectral library and the dataset acquired by Lesaignoux et al. (2013). The best relations were obtained between reflectances in the mid-infrared channels (ETM5 and ETM7) and the thermal infrared channel (ETM6) with correlation coefficients of 0.63 and 0.72 respectively. The relations were mostly generated by the variations of soil reflectances due to changes in soil moisture. Correlations were lower when considering the variations due to soil type.

Mineralogical clays are intimately mixed in soils. Until now, there is not a clear view on which unmixing methods are the most effective in case of intimate mixtures with both clays and with other mineral. A laboratory approach is proposed in order to compare existing linear and nonlinear unmixing approaches in terms of performances. Laboratory images of pure clays of Montmorillonite, Illite and Kaolinite were preprocessed using 6 transformations commonly used in literature (SNV, CR, CWT, Hapke, 1stSGD, Log(l/R)). Prediction of abundances has been performed on non-preprocessed and preprocessed data with 2 linear unmixing methods (FCLS and MESMA) and 2 nonlinear unmixing methods (GBM and MLM). Results show a decrease of unmixing performance due to sample variability, which can be reduced using the appropriate preprocessing. However, no particular advantage has been found on using nonlinear over linear unmixing approaches for clay minerals.

Hyperspectral remote sensing data offer unique opportunities for the characterization of land surface in urban areas. However, no hyperspectral- unmixing based studies have been conducted to automatically detect photovoltaic panels, which represent one of the important components of energy systems in such areas. In this paper, a hyperspectral-unmixing based method is proposed to detect photovoltaic panels and to estimate their areas. This approach is based on an original multiplicative nonnegative matrix factorization (NMF) algorithm with some known photovoltaic panel spectra. The proposed method can be considered as a partial/informed NMF approach. Experiments are conducted on realistic synthetic and real data to evaluate the performance of the proposed approach. In both cases, obtained results show that the proposed method yields much better overall performance than a method from the literature.

The monitoring of the water cycle at the Earth surface which tightly interacts with the climate change processes as well as a number of practical applications (agriculture, soil and water quality assessment, irrigation and water resource management, etc…) requires surface temperature measurements at local scale. Such is the goal of the Indian-French high spatio-temporal TRISHNA mission (Thermal infraRed Imaging Satellite for High-resolution Natural resource Assessment). The scientific objectives of the mission and research work conducted to consolidate the mission specifications are presented. Progress in modelling of surface fluxes is then discussed. The main specifications of the mission such as the revisit, the spatial resolution, the overpass time, the spectral bands and the orbit are analyzed and justified. The resulting baseline of the mission is given.

A strong intern dynamic characterizes towns, a very high spatial heterogeneity of their elements, their 3D geometric shapes (horizontal and vertical) inducing shadows, and their large variety of materials. These characteristics make the collection of information of land surface properties and urban descriptors more delicate. Due to the enhancement of spatial to deepen the observation of urban areas. Nevertheless, such a type of sensors would not contribute to the characterization of the urban land surface properties (chemical composition of materials, species of vegetation, quality of soils, etc.). They and show great potentials might consider Hyperspectral imagery capacities as providing useful products but it becomes mandatory to define which type of information these different sensors can deliver. The ANR HYEP project has the purpose to demonstrate the benefit of a second generation of hyperspectral space borne mission characterized by a high spatial resolution (8m GSD) and a high temporal revisit. After a detailed description of the motivation of such a proposal, applications are given focused on urban vegetation, sealed and impervious areas, solar panel area estimation.

This study aims at evaluating two classes of methods to discriminate 13 peatland vegetation types using reflectance data from hyperspectral in situ measurements. These vegetation types were empirically defined according to their composition, strata and biodiversity richness. We suppose that specific biophysical properties/components may help discriminating vegetation types applying supervised classification such as Random Forest (RF), Support Vector Machines (SVM), Regularized Logistic Regression (RLR), Partial Least Squares-Discriminant Analysis (PLS-DA). Biophysical components can be used in a local way considering vegetation spectral indices or in a global way considering spectral ranges which characterize specific biophysical properties. and transformed spectral signatures enhancing absorption features. The results of this study suggest that RLR classifier is promising to map the different vegetation types with high ecological values despite vegetation heterogeneity and mixture.

Hyperspectral imagery generates huge data volumes, consisting of hundreds of contiguous and often highly redundant spectral bands. Difficulties are caused by this high dimensionality. Feature selection (FS) is a possible strategy to reduce the number of bands, consisting in selecting the most relevant bands for a classification problem. It is adapted to the design of superspectral sensor dedicated to specific applications. FS is an optimization problem involving both a metric (that is to say a FS score or criterion measuring the relevance of feature subsets) to optimize and an optimization strategy. In this paper, a wrapper FS score based on Random Forests (RF) and taking into account RF class membership measures was proposed. It was compared to a state-of-the-art wrapper FS score (classification Kappa obtained by RF). Both were then evaluated quantitatively considering both classification performance reached applying different classifiers. An qualitative analysis was also performed to consider the stability/regularity of the selected features along the spectrum. Even though the quantitative evaluation showed little differences between the two tested FS criteria, there seemed to be a trend in favour of the proposed criterion. Taking into account the measures of class membership provided by a RF classifier slightly improved results, regularizing feature selection.

This study aims at producing total petroleum hydrocarbon content map from airborne hyperspectral images of bare soils in the 0.4-2.5μm spectral domain. Existing methods, such as Kühn's Hydrocarbon Index or Allen & Satterwhite's Normalized Difference Hydrocarbon Index, are not able to produce reliable quantitative maps. To this end, more elaborated methods have recently been developed out of laboratories sample spectra. They use multivariate methods to build predictive models from a population of samples observed in controlled laboratory conditions. However, this laboratory models have not been extended to field nor airborne hyperspectral data. This paper presents the first results from simultaneous field and airborne measurements above sand/HC mixture samples.

This paper presents a new algorithm for the analysis of linear spectral mixtures in the thermal infrared domain, with the goal to jointly estimate the abundance and the subpixel temperature in a mixed pixel, i.e., to estimate the relative proportion and the temperature of each material composing the mixed pixel. This novel approach is a two-step procedure. First, it estimates the emissivity and the temperature over pure pixels using the standard temperature and emissivity separation (TES) algorithm. Second, it estimates the abundance and the subpixel temperature using a new unmixing physics-based model, called Thermal Remote sensing Unmixing for Subpixel Temperature (TRUST). This model is based on an estimator of the subpixel temperature obtained by linearizing the black body law around the mean temperature of each material. The abundance is then retrieved by minimizing the reconstruction error with the estimation of the subpixel temperatures. The TRUST method is benchmarked on simulated scenes against the fully constrained least squares unmixing applied on the radiance and on the estimation of surface emissivity using the TES algorithm. The TRUST method shows better results on pure and mixed pixels composed of two materials. TRUST also shows promising results when applied on thermal hyperspectral data acquired with the Thermal Airborne Spectrographic Imager during the Detection in Urban scenario using Combined Airborne imaging Sensors campaign and estimates coherent localization of mixed-pixel areas.

The estimation of surface emissivity and temperature from thermal hyperspectral data is a challenge. Methods that estimate the temperature and emissivity on a pixel composed by one single material exist. However, the estimation of the temperature on a mixed pixel, i.e. a pixel composed by more than one material, is more complex and has scarcely been investigated in the literature. This paper addresses this issue by proposing an estimator which linearizes the Black Body law around the mean temperature of each material. The performance of this estimator is studied using simulated data with different hyperspectral sensor configurations and under various noise conditions. The obtained results are encouraging and show an accuracy on the estimated temperature of 0.5 K while using high spectral resolution sensor.

The estimation of surface emissivity and temperature from thermal hyperspectral data is a challenge. There are several methods that estimate the temperature and the emissivity by assuming that the pixel is composed by a single material. However, the estimation of the temperature on a mixed pixel, i.e. a pixel composed by more than one material, is more complex and has scarcely been investigated in the literature. This paper addresses this issue by jointly estimating the materials composing the mixed pixel and their temperatures. It uses an unmixing method based on the linearization of the Black Body law. The performance of this strategy is studied using synthetic data and a real thermal image acquired by the TASI sensor.

A possible step in dimensioning future space-based full-waveform lidar sensors is to predict space signals from commercial airborne laser scanner data. This method has proved able to simulate passive satellite sensors with precise accounting of the scene heterogeneity effects. In this letter, we use the DELiS code (n-Dimensional Estimation of Lidar Signals) to numerically evaluate a simple, efficient aggregation method for combining airborne lidar measurements (submeter footprints) into space lidar signals (decametric footprints). Two main sources of error are studied: the heterogeneity of the scene combined with an insufficient coverage by the airborne scanner, and the multiple scattering of the laser pulse in vegetation. It is found that for three different types of vegetation (corn, orchard, rainforest), and in three usual scanning configurations, the satellite signal can be derived with good precision. However, multiple scattering in the vegetation is shown to induce errors of up to 30% of the total backscattered signal depending on the wavelength.

In the field of remote sensing, the unmixing of hyperspectral images is usually based on the use of a mixing model. Most existing spectral unmixing methods, used in the reflective range (0.4–2.5$\mu\hbox{m}$), rely on a linear model of endmember reflectances. Nevertheless, such a model supposes the pixels at the ground level to be uniformly irradiated and the scene to be flat. When considering a 3-D landscape, such a model is no longer valid as irradiated and shadowed areas are present, as well as radiative interactions between facing surfaces. This paper introduces a new mixing model adapted to urban environments and which aims to overcome these limitations. This model is derived from physical equations based on radiative transfer theory, and its analytic expression is linear–quadratic. Similar models have already been used in the literature for unmixing purposes but without being justified by physical analysis. Our proposed model is validated using a synthetic but realistic European 3-D urban scene. Then, simplifications are introduced, based on a study of the different radiative components contributing to the signal in a way to make the model easy to use for spectral unmixing. This paper also shows that the quadratic term cannot be neglected in many cases in urban environments since it can, e.g., range from 15% to 20% of the reflectances in canyons.

The temperature and emissivity separation (TES) algorithm is used to retrieve the land surface emissivity (LSE) and land surface temperature (LST) values from multispectral thermal infrared sensors. In this paper, we analyze the performance of this methodology over urban areas, which are characterized by a large number of different surface materials, a variability in the lowest layer of the atmospheric profiles, and a 3-D structure. These specificities induce errors in the LSE and LST retrieval, which should be quantified. With this aim, the efficiency of the TES algorithm over urban materials, the atmospheric correction, and the impact of the 3-D architecture of urban scenes are analyzed. The method is based on the use of a 3-D radiative transfer tool, TITAN, for modeling all of the radiative components of the signal registered by a sensor. From the sensor radiance, an atmosphere compensation process is applied, followed by a TES methodology that considers the observed scene to be a flat surface. Finally, the retrieved LSE and LST are compared with the original parameters. Results show the following: First, the TES algorithm used reproduces the LSE (LST) of urban materials within a root-mean-square error (rmse) of 0.017 (0.9 K). Second, 20% of uncertainty in the water vapor content of the total atmosphere introduces an rmse of 0.005 (0.4 K) for the LSE (LST) product. Third, in a standard case, the 3-D structure of an urban canyon leads to an rmse of 0.005 (0.2 K) for the LSE (LST) retrieval of the asphalt at the bottom of the scene.

This letter describes a laboratory experiment where reflectance signatures of three plant species are measured at a leaf level in the 3-15-μm spectral domain. The leaf samples are progressively dried in order to analyze the behavior of their spectral signature according to the variations in their water content. Our first objective aims to underline leaf water content (LWC) impact on the spectral signatures. This work is a necessary step toward further studies dealing with interpretation of multispectral remote sensing data or estimation of water stress and energy budget. The drying process and measurement method are detailed. This letter deals with dry and fresh leaves (as found in literature) and considers intermediate water content levels as well. For intermediate LWC levels, our analysis outlines some important results: The spectral domain may be divided into two parts, namely, 3-5.5 and 5.5-15 μm, each corresponding to different impacts of LWC variation; both sides of a cherry tree leaf have not the same behavior according to the water content amount; and, in the 8-15-μm, the drying process impacts when the LWC becomes lower than a threshold value (around 30%).

The paper presents a new method of image simulation. More precisely, it focus on the most important step of this method. This step consists in retrieving reflectance composing an urban scene thanks to a set of multi-angular acquisitions. The high resolution of these acquisitions make the problem complex and the “flat ground” methods become obsolete. The new developed method takes into account the 3D geometry of the scene and the directional effects of urban material. It is first described and, then is applied to real data, aerial images over Toulouse.

The purpose of this paper is the capacity of spectral reflectance, in the full optical domain 0.4 - 12 μm, to retrieve the Soil Moisture Content (SMC) of bare soils. For this, a new empirical soil model was performed from spectra lab measurements, and more precisely from soil's classification, to simulate spectral reflectance for a given SMC, and in inverse way to estimate SMC. In solar and thermal domain, simulated spectra are highly correlated with measured spectra, respectively with 98% and 85%. In inverse way, estimation of SMC from a spectral reflectance and a given soil classes, provides a RMSE of 3.2% in solar and 2.8% in thermal domain.

The purpose of this paper is to analyse the impact of surface soil moisture on spectral reflectance in the optical domain [0.4-15 μm]. This work is based on lab spectral reflectance measurements of many bare soils at different moisture contents. Firstly, a classification of bare soil samples is performed according to their spectral signatures: five classes are then defined. Secondly, the soil moisture content impact on spectral signatures is analysed. In the [0.4-15 μm] domain, measurements exhibit, for all the samples, a decreasing of the reflectance level with an increasing of moisture content. These measurements give information on absorption peaks related to soil mineral components like hydroxyl, carbonate, and quartz. Thus, analysis of our lab measurements indicates that soil moisture impact on spectral reflectance depends of studied spectral domain. These measurements may improve existing data bases, and will be used in a processing chain to estimate the soil moisture content (cases of bare soil and/or sparse vegetation) in the optical domain [0.4-12 μm] by using airborne hyperspectral imaging.

A semianalytical model, named APOM (aerosol plume optical model) and predicting the radiative effects of aerosol plumes in the spectral range [0.4,2.5 mum], is presented in the case of nadir viewing. The scene is represented by an atmospheric layer (molecules and natural aerosols) located above the plume layer. The estimated at-sensor reflectance depends on the solar zenith angle, the plume optical properties (optical depth, single-scattering albedo and asymmetry parameter), the ground reflectance and the wavelength. Its numerical coefficients are derived from COMANCHE radiative transfer simulations. Model accuracy is assessed by using a set of simulations performed in the case of biomass burning and industrial plumes. APOM proves to be accurate and robust for solar zenith angles between 0deg and 60deg whatever the sensor altitude, the standard atmosphere and for plume phase functions defined from urban and rural models. The modeling errors in the at-sensor reflectance are on average below 0.002. They can reach values as high as 0.01 for wavelengths close to 0.4 mum but mainly correspond in such cases to low relative errors (below 5% and 3% on average). This model can be used for forward modeling (quick simulations of multi/hyperspectral images, help in sensor design...) as well as for the retrieval of the plume optical properties. For this purpose, we propose a model for the optical properties involving a few parameters and we show that even in this case the problem is still ill-constrained and that constraints have to be imposed. First retrieval results as well as recommendations for the inversion are presented.

Two reciprocal direct and inverse radiative transfer models dealing with hyperspectral remote sensing in the visible-to-shortwave-infrared spectral domain are described in this paper. The first one, called COMANCHE, considers a flat and heterogeneous ground scene, with bidirectional reflectance effects, and computes spectral radiance hypercubes at the sensor level. Trapping and environment phenomena are take into account through specific optimized Monte Carlo modules. The reciprocal inverse algorithm, called COCHISE, considers a sensor-level hyperspectral image and retrieves the ground spectral reflectance distribution as well as the water vapor content. COCHISE removes the atmospheric and environment effects with the same modeling as COMANCHE, but consider however the Lambertian assumption for the ground reflectance. Both of the models are validated with existing radiative transfer codes (MODTRAN and AMARTIS, for instance), and also using experimental datasets from the Airborne Visible/Infrared Imaging Spectrometer. The comparisons show very good agreement regarding to the usual uncertainties involved insuche experiment. COCHISE is also applied on a additional dataset acquired by HyMap.

The issue of deriving cross-scale aggregation rules has been extensively investigated over the last two decades. A widely used approach consists of formulating grid-scale surface radiances using the same equations that govern the patch-scale behavior but whose arguments are the aggregate expressions of those at the patch-scale. This approach derives the area-averaged or effective radiative surface temperature as might be observed using low spatial resolution satellite data. The problem however is that such satellite data exhibit large directional effects and no study has addressed this issue. The present work tackles this problem in the thermal infrared domain. The directional effects are studied by modeling. Thus, an infrared sensor observing a two-dimensional (2-D) heterogeneous plane surface is modeled. The 2-D heterogeneous plane surface is simulated by a grid with two homogeneous elements (vegetation-bare soil). The angular properties of the local surfaces, assumed homogeneous, are calculated by a multiple scattering model. The equivalent angular radiance of the complete heterogeneous scene is then determined by applying the aggregation method. This radiance is very sensitive to the surface heterogeneity, especially when the spatial variation of the surface temperature is significant and when the directional behavior of the surface is non-Lambertian. As a result, an angular variation of 6% on radiance was obtained on a heterogenous surface between a zenith angle of 70/spl deg/ and on-nadir measurements.

This paper analyzes the signal measured by optical remote sensors when acquiring data over a shaded part of an urban scene. The signal is much lower for this kind of target than for others because there is no direct downward irradiance. Here, a simple urban scene is considered with a shaded area. The signal observed by a high spatial resolution satellite sensor over an ordinary panchromatic band (500-700 nm) is computed thanks to a radiative transfer code [advanced modeling of the atmospheric radiative transfer for inhomogeneous surfaces (Amartis)] capable of dealing with ground topography and heterogeneity. The signal is analyzed, and it appears that environmental effects play a significant role. Moreover, because of the scattering that occurs at shorter wavelengths, it is also shown that a widening of the band to 440 nm sharpens the difference between signals coming from two different ground types (for whose the difference of reflectance is constant and equal to 0.1) by about 10%. This demonstrates that the band widening may be beneficial to observation in shadow, mainly because of scattering effects. A more realistic scene is also considered, in which each part is associated with realistic spectral properties. This simply shows the importance of the thematic in the choice of band, as it determines the effect of the widening.

Pixel fusion is used to elaborate a classification method at pixel level. It needs to take into account the as accurate as possible information and take advantage of the statistical learning of the previous measurements acquired by sensors. The classical probabilistic fusion methods lack performance when the previous learning is not representative of the real measurements provided by sensors. The Dempster-Shafer theory is then introduced to face this disadvantage by integrating further information which is the context of the sensor acquisitions. In this paper, we propose a formalism of modeling of the sensor reliability in the context that leads to two methods of integration: the first one amounts to integrate this further information in the fusion rule as degrees of trust and the second models the sensor reliability directly as mass function. These two methods are compared in the case where the sensor reliability depends on an atmospheric disturbance: the water vapor.

The application of multi-sensor fusion, which aims at recogizing a state among a set of hypotheses for object classification, is of major interest as regards the performance improvement brought by the sensor complementarily. Nevertheless this needs to take into account the more accurate as possible information and take advantage of the statistical learning of the previous measurements acquired by sensors. The classical probabilistic fusion methods lack of performance when the previous learning is not representative of the real measurements provided by sensors. The theory of evidence is then introduced to face this disadvantage by integrating a further information which is the context of the sensor acquisitions. In this paper, we propose a formalism of modeling of the sensor reliability to the context that leads to two methods of integration when all the hypotheses, associated to the objects of the scene acquired by sensors, are previously learnt: the first one amounts to integrate this further information in the fusion rule as degrees of trust and the second models the sensor reliability directly as mass functions. These two methods are based on the theory of fuzzy events. Afterwards, we are interested in the evolvement of these two methods in the case where the previous learning is unavailable for a hypothesis associated to an object of the scene and compare these two methods in order to deduce a global method of contextual information integration in the fusion process.

POLDER is a CNES instrument on board NASDA's ADEOS polar orbiting satellite, which was successfully launched in August 1996. On October 30, 1996, POLDER entered its nominal acquisition phase and worked perfectly until ADEOS's early end of service on June 30, 1997. POLDER is a multispectral imaging radiometer/polarimeter designed to collect global and repetitive observations of the solar radiation reflected by the Earth/atmosphere system, with a wide field of view (2400 km) and a moderate geometric resolution (6 km). The instrument concept is based on telecentric optics, on a rotating wheel carrying 15 spectral filters and polarizers, and on a bidimensional charge coupled device (CCD) detector array. In addition to the classical measurement and mapping characteristics of a narrow-band imaging radiometer, POLDER has a unique ability to measure polarized reflectances using three polarizers (for three of its eight spectral bands, 443 to 910 nm) and to observe target reflectances from 13 different viewing directions during a single satellite pass. One of POLDER's original features is that its in-flight radiometric calibration does not rely on any on-board device. Many calibration methods using well-characterized calibration targets have been developed to achieve a very high calibration accuracy. This paper presents the various methods implemented in the in-flight calibration plan and the results obtained during the instrument calibration phase: absolute calibration over molecular scattering, interband calibration over sunglint and clouds, multiangular calibration over deserts and clouds, intercalibration with Ocean Color and Temperature Scanner (OCTS), and water vapor channels calibration over sunglint using meteorological analysis. A brief description of the algorithm and of the performances of each method is given.