Aichouche Belhadj-Aissa

InSAR Phase Filtering via Joint Subspace Projection Method: Application in Change Detection

A new approach is presented to estimate the synthetic aperture radar (SAR) interferometric phase via the joint subspace projection. In this letter, a new formulation is proposed to transform the cost function minimization problem into the problem of finding the roots of a second-order polynomial whose coefficients are evaluated via the joint noise subspace. The proposed method enjoys a substantially reduced computational complexity and improves the estimation of the interferometric phase. Results obtained using real SAR data show the effectiveness of the proposed method. In the change detection application, the method offers much better separation between the changed and unchanged coherence pixels than existing filters.

Statistical Similarity Measure for Oil Slick Detection in SAR Image

Spaceborne Synthetic Aperture Radar (SAR) is well adapted to detect ocean pollution independently from daily or weather condition. As it is sensitive to surface roughness, the presence of oil film on the sea surface decreases the backscattering of the sea surface resulting in a dark feature patches in SAR images. In fact, oil slicks have specific impact on ocean wave spectra. Initial wave spectra may be characterized by three kinds of waves, big, medium and small, which correspond physically to gravity and gravity-capillary waves. The increase of viscosity due to the presence of oil damps gravity-capillary waves. This induces a damping of the backscattering to the sensor, but also a damping of the energy of the wave spectra, then it modifies the sea surface roughness observed by the sensor. Thus, local detection of wave spectra modification may be achieved by a appropriated texture analysis of the original SAR image. In this paper, the texture analysis is based on measure of similarity between a local probability density function (pdf) of clean water and the local pdf of the zone to be inspected. The local distribution is estimated in the neighbourhood of each pixel, through a sliding window, and compared to the reference one by using the Kullback-Leibler (KL) distance between distributions. An efficient strategy has been adopted in order to perform pdf estimation through a non-parametric approach.

Multi-scale segmentation for remote sensing imagery based on minimum heterogeneity rule

Image segmentation is an essential step toward higher level image processing in remote sensing. However, the traditional image segmentation approaches based on pixels spectral characteristics and single-scale image information extraction methods have obvious flaws in this respect. Currently, multi-scale image segmentation is seen as a promising alternative of traditional segmentation method and is one of the most useful approaches in object oriented classification of remotely sensed images. In this paper, we present a multi-scale segmentation method based on Minimum Heterogeneity Rule (MHR) for merging objects. Segmentation results show that this method can easily adapt its scale parameter to different scale image analysis tasks and any chosen scale object-extraction of interest.

Sea SAR Images Analysis to Detect Oil Slicks in Algerian Coasts

In this paper, we investigate the performance of partition features derived from histogram analysis to isolate dark spots which are candidates to be oil spills in SAR images. The first partition is carried out to obtain preliminary clusters of the pixels on the basis of their grey level intensities and threshold values deduced from the histogram. The detection process is achieved by a contextual partition where the conflict pixels are attributed to their region involving local information about pre-etiqueted pixels neighbouring the pixel in question. For pixel’s assignment, we propose two decision criteria: the first based on Local Probability Maximization (LPM) while the second uses a Chi-squared test (χ2). We considered variable context in order to characterize the sea texture and dark spots. This method is tested on ERS-2 SAR Precision Image (PRI) covering Algerian coasts and gave promising results which are useful for the identification process.

Improving Ccd Performance by the Use of Local Fringe Frequencies

Coherent Change Detection (CCD) using multi-temporal Synthetic Aperture Radar (SAR) is one of the most important applications of remote sensing technology. With the advent of high- resolution SAR images, CCD has received a lot of attention. In CCD, the interferometric coherence between two SAR images is evaluated and analyzed to detect surface changes. Unfortunately, the sample coherence estimator is biased, especially for low-coherence values. The consequence of this bias is the apparition of highly coherent pixels inside the changed area. Within this context, the detection performance will considerably degrade, particularly when using high resolution SAR data. In this paper, we propose a new CCD method based on cleaning of coherence inside changed areas, which is characterized by high Local Fringe Frequencies (LFF) values, followed by a space-averaged coherence method. According to the proposed method, the results obtained with Cosmo-SkyMed (CSK) SAR data show an enhancement of change detection performance of about 6% while preserving subtle changes.

Unsupervised Multispectral Image Classification using Artificial Ants

Based on the existing works dealing on data clustering with artificial ants, we contribute in this paper to resolve a real clustering problem related on unsupervised multispectral image classification using ants approach, where classes are found without the a priori knowledge of the correct number of classes. Knowing that most of the unsupervised classification methods require the definition of a probable number of classes and an initial partition, the proposed ant-based approach is very interesting insofar for remotely sensed data over the whole of earth, it is not easy to obtain this a priori knowledge

Multi-scale Satellite Images Fusion using Dempster Shafer Theory

The general principle of the multi-scale fusion is based on modelling of mixed pixels of the low resolution image and confuses classes of the high resolution image. This model involves taking into account not only the simple classes distinguishable at the two resolutions, but also union of classes not distinguishable through the resolutions. We present in this paper a work which focuses on the realization of a multi-scale images fusion process based on the Dempster-Shafer evidence theory (DST). The aim of the fusion process is the improvement of the land cover map by exploiting the rich spectral information of low spatial resolution images and the rich spatial information of high spatial resolution images. The multi-scale fusion using unsupervised model and supervised model allows generating new spectral classes indistinguishable on the high and low spatial resolution, thus obtaining land cover map rich in spectral and spatial information. However, if we have prior knowledge about the searched fusion classes, the application of the supervised model is recommended, otherwise it is preferable to apply the unsupervised model for all possible fusion classes, generated by the mathematical formalism of DST.

Improvement of Land Cover Map from Satellite Imagery using DST and DSmT

The aim of this paper is to show that Dempster-Shafer theory (DST) and a recent theory of plausible and paradoxical reasoning introduced by Dezert and Smaradache and thus called Dezert-Smarandache theory (DSmT), can be successfully applied to improve a supervised classification of remotely sensed data. Notice that application fields of these two theories are related on multisensor/multitemporal/multiscale data fusion. In this study, our contribution lies in developing a new multispectral data classification process which can be seen as a multisensor fusion process where each thematic class is considered as one source of information

Supervised classification of remotely sensed images using Bayesian network models and Kruskal algorithm

The aim of this work is to evaluate the performances of three Bayesian networks widely used for supervised image classification. The developed structures are constructed due to Kruskal algorithm which allows the determination of the maximum weight spanning tree by using the mutual information between the attributes. We started by the Bayesian naïve classifier (BNC), which assumes that there is no dependency, between the attributes to classify. In order to relax this strong assumption, we tested the tree augmented naïve Bayes classifier (TANC) where each feature has at most one variable as parent, and the forest augmented naïve Bayes classifier (FANC) where each attribute forms an arbitrary graph rather than just a tree. These classifiers are evaluated using a multispectral image and hyperspectral image in order to analyze the structure classifier complexity according to the number of attributes (04 and 10 spectral bands for the two images respectively). Obtained results are compared with state-of-the art competitor, namely, the SVM classifier. Classified images by TANC and FANC achieved higher accuracies than other classifiers including SVM. It is concluded that the choice of attributes dependencies significantly contributes to the discrimination of subjects on the ground. Thus, Bayesian networks appear as powerful tool for multispectral and hyperspectral image classification.

Object-oriented SVM classifier for ALSAT-2A high spatial resolution imagery: A case study of algiers urban area

The development of robust object-oriented classification approaches suitable for medium to high spatial resolution satellite imagery provides a valid alternative to traditional pixel-based classification approaches. In the past, Support Vector Machines (SVM) have been tested and evaluated only as pixel-based image classifiers. Moving from pixel-based analysis to object-based analysis, a fuzzy classification concept is generally used through eCognition software [1]. In this paper, we propose an object-oriented classification system based on SVM approach. By using a suitable scale during a multi-resolution segmentation step, obtained results are compared to those produced by a pixel-based SVM classifier. The classification process is performed by using a high spatial resolution imagery acquired by the Algerian satellite ALSAT-2A. From the comparison of obtained results, it is concluded that the object-based classifier is more efficient than the pixel-based classifier for the discrimination of seven major land cover classes.