Search

We introduce a normalization algorithm which highlights short-term, localized, non-periodic fluctuations in hyper-temporal satellite data by dividing each pixel by the mean value of its spatial neighbourhood set. In this way we suppress signal patterns that are common in the central and surrounding pixels, utilizing both spatial and temporal information at different scales. We test the method on two subsets of a hyper-temporal thermal infra-red (TIR) dataset. Both subsets are acquired from the SEVIRI instrument onboard the Meteosat-9 geostationary satellite; they cover areas with different spatiotemporal TIR variability. We impose artificial fluctuations on the original data and apply a window-based technique to retrieve them from the normalized time series. We show that localized short-term fluctuations as low as 2K, which were obscured by large-scale variable patterns, can be retrieved in the normalized time series. Sensitivity of retrieval is determined by the intrinsic variability of the normalized TIR signal and by the amount of missing values in the dataset. Finally, we compare our approach with widely used techniques of statistical and spectral analysis and we discuss the improvements introduced by our method. HighlightsWe introduce a normalization approach for detection of extremes.We consider both the spatial and temporal dimensions of geophysical data.We apply the method and test its sensitivity on hyper-temporal satellite data.

A graphics processing unit (GPU) can perform massively parallel computations at relatively low cost. Software interfaces like NVIDIA CUDA allow for General Purpose computing on a GPU (GPGPU). Wrappers of the CUDA libraries for higher-level programming languages such as MATLAB and IDL allow its use in image processing. In this paper, we implement GPGPU in IDL with two distance measures frequently used in image classification, Euclidean distance and spectral angle, and apply these to hyperspectral imagery. First we vary the data volume of a synthetic dataset by changing the number of image pixels, spectral bands and classification endmembers to determine speed-up and to find the smallest data volume that would still benefit from using graphics hardware. Then we process real datasets that are too large to fit in the GPU memory, and study the effect of resulting extra data transfers on computing performance. We show that our GPU algorithms outperform the same algorithms for a central processor unit (CPU), that a significant speed-up can already be obtained on relatively small datasets, and that data transfers in large datasets do not significantly influence performance. Given that no specific knowledge on parallel computing is required for this implementation, remote sensing scientists should now be able to implement and use GPGPU for their data analysis.

A common challenge in remote sensing is the classification of objects that are spectrally similar but represent physically different types of ground cover. In this paper, we describe and apply three complementary shape measures to classify morphologically different waterbodies in a Landsat image. Image segmentation was used to create objects of image pixels containing water, and shape measures were calculated for all obtained objects. A shape-based, a spectra-based and a combined spatial-spectral classification were carried out on a subset of the image using endmembers acquired outside the subset. The spectral classification was based on Euclidean distance. The shape-based and combined spectral-shape classification were based on vector angle, as the chosen shape measures are influenced by the image lattice and could only be used as a relative measure. The results of this approach are discussed and compared to an expert interpretation of the same dataset. Results show that shape measures are affected by image resolution and should be used as a relative measure when objects consist of 500 pixels or less. Although the combined spectral-shape classification was not satisfactory and needs more research, the classification that is solely based on shape measures can distinguish spectrally identical waterbodies and had a score of 94% compared to the expert classification.

Edge operators are widely used on gray-level images and are recently improved to work with multispectral and even hyperspectral imagery. The high spectral information content in hyperspectral images allows a detailed description of boundaries and thus a supervised boundary detection. In this letter, we describe a template matching algorithm for the detection of fuzzy and crisp boundaries. For this purpose, the template has a one-dimensional design consisting of two different spectra. This template is matched to a remote sensing image by moving and rotating the template over the image. A statistical spatial and spectral fit of the template is calculated for every position and orientation. Important steps in this approach are the design of a template according to our knowledge of a boundary, and, mainly depending on the template design, the interpretation of the algorithm output. The algorithm has been used for the detection of boundaries between selected mineral assemblages in a hyperspectral image that covers a hydrothermal alteration system. Results show that the algorithm successfully detects the boundaries that had been defined in the templates. In addition, it is shown that rotation of the template in the algorithm reveals information on the type of boundary (crisp or fuzzy) and identifies pixels where only one of the template endmembers is present

Most endmember extraction algorithms are based on the spectral properties of the dataset only to discriminate between the pixels. Endmembers with distinct spectral profiles or high spectral contrast are easier to detect, whereas the endmembers having low spectral contrast with respect to the whole image are difficult to determine. The spatial-spectral integration approach searches for endmembers by analyzing the image in subsets such that it increases the local spectral contrast of the low contrast endmembers and increases their odds of selection. Spatial spectral integration process utilizes Hyperspectral subspace identification by minimum error (HySime) to determine a set of locally defined eigenvectors explaining the maximum variability of the subsets of the image. The image data is then projected onto these locally defined eigenvectors which produces a set of candidate endmember pixels. The candidate endmember pixels, that are spectrally similar and having similar spatial coordinates, are averaged together and grouped into different endmember classes. The method is applied to spaceborne hyperspectral dataset to illustrate the effects of using spatial measures in the process of endmember extraction. The spatial-spectral integration results show that the endmember pixels obtained by imposing spatial constraints are cleaner and more representative of the land use land cover classes.

The OMEGA instrument onboard of ESA's Mars Express mission is the first hyperspectral sensor that has collected data from Mars. The OMEGA team has shown that Mars has considerable surface compositional variation. Spectral interpretation and mineral mapping, however, is difficult on a pixel-by-pixel basis due to sensor noise, an atmosphere dominated by carbon dioxide and especially an unknown surface cover. An object- based segmentation approach is for datasets that are acquired in areas from which we do not have a-priori knowledge useful to ignore the scene-wide effects of the unknown atmosphere and to enhance the spectral contrast of the planet's surface, without any human bias. Unlike common segmentation procedures where distances in feature space are used for pixel similarity criteria, the OMEGA data is segmented using similarity criteria based on spectral absorption feature parameters such as position, depth and area. This paper shows the first results of an object-based processing of OMEGA data and discusses possibilities of future development.

This study presents a novel approach in combining textural information from the panchromatic band with spectral information from the multispectral bands for improved image classification. Firstly, we develop a texture measure based on wavelet coefficients of the panchromatic band that can be aggregated to the resolution of the multispectral bands. Secondly, we combine the texture measures with the spectral information in the multispectral bands in a fuzzy classification framework. Thirdly, we illustrate our approach with a case study of vegetation and land cover classification based on a Quickbird image of subantarctic Macquarie Island.