Detecting onshore hydrocarbon is a major topic for both environmental monitoring and exploration. In this work, a hyperspectral image acquired nearby an old oil extraction site in tropical area is analyzed. The area of interest includes a pit filled with bio-degraded heavy oil, surrounded by herbaceous vegetation and many lagoons. <p> </p>First, we focused on methodologies that can detect oil pollution in an unsupervised manner. Based on the assumption that such oil pits are rare events in the image, statistical approach for anomalies detection, derived from the Reed-Xiaoli detector, is used. In order to decrease the number false alarms, some a priori knowledge about the spectral signature of the pits and about the background is introduced. This approach succeeds in detecting the pit with very few false alarms. <p> </p>Hydrocarbon pollution can have an impact on vegetation and leads to change in vegetation (bio)physical parameters (pigments, water content, …), according to species, pollutant type and exposition time . In order to map the polluted area without any a priori knowledge, several un-supervised classification, including an original method of automatic classification combining unmixing approach and SVM (support Vector Machine) are applied and compared. The results are compared with a partial “ground truth map” that has been derived from visual observations on the field, and with areas of stressed vegetation that have been mapped using combination of specific spectral indices. The classification results are consistent with the ground truth map and the retrieved stressed vegetation areas.
This paper presents an unsupervised segmentation method dedicated to vegetation scenes with decametric or metric spatial resolutions. The proposed algorithm, named SIEMS, is based on the iterative use of the Expectation–Maximization algorithm and offers a good trade-off between oversegmentation and undersegmentation. Moreover, the choice of its input parameters is not image–dependent on the contrary to existing technics and its performances are not crucially determined by these input parameters. SIEMS consists in creating a coarse segmentation of the image by applying an edge detection method (typically the Canny–Deriche algorithm) and splitting iteratively the undersegmented areas with the Expectation–Maximization algorithm. The method has been applied on two images and shows satisfactory results. It notably allows to distinguish segments with slight radiometric variations without leading to oversegmentation.
In this paper, we propose an innovative classification method dedicated to hyperspectral images which uses
both spectral information (Principal Component Analysis bands, Minimum Noise Fraction bands) and spatial
information (textural features and segmentation). The process includes a segmentation as a pre-processing step,
a spatial/spectral features calculation step and finally an area-wise classification. The segmentation, a region
growing method, is processed according to a criterion called J-image which avoids the risks of over-segmentation
by considering the homogeneity of an area at a textural level as well as a spectral level. Then several textural
and spectral features are calculated for each area of the segmentation map and these areas are classified with
a hierarchical ascendant classification. The method has been applied on several data sets and compared to the
Gaussian Mixture Model classification. The JSEG classification process finally appeared to gives equivalent, and
most of the time more accurate classification results.
In this paper, variations with wavelength of aerosol optical properties which are optical thickness τ, single-scattering albedo ω<sub>0</sub> and asymmetry parameter g are modeled using polynomial functions in the case of dense biomass burning plumes in the spectral range [0.4 - 1.1 μm]. Optical properties are computed from Mie theory for various types of particles, size distributions and concentrations. In a first step, each optical property is fitted by polynomials with one, two and three parameters over the whole set of optical properties and then an error analysis is performed in order to choose the optimal number of parameters depending on wished accuracy. In a second step, the impact of modeling errors on top of atmosphere reflectance ρ<sup>TOA</sup> is investigated depending on ground reflectance. The impact on ρ<sup>TOA</sup> of ground reflectance variability under the smoke plume is also assessed. Calculations show that accurate modeling of spectral behaviour requires three parameters for τ and ω<sub>0</sub> and two parameters for <i>g</i>. It leads to simulations of ρ<sup>TOA</sup> with an accuracy of about 0.001 which is compatible with the level of noise of current sensors. Using one less parameter for each optical property yields errors on ρ<sup>TOA</sup> within 0.02.