Luigi Mascolo

Iterative Hyperspectral Image Classification Using Spectral–Spatial Relational Features

This paper describes the principles and implementation of an algorithm for the classification of hyperspectral remote sensing images. The proposed approach is novel and can be included within the category of the spectral-spatial classification algorithms. The elements of novelty of the algorithm are as follows: (1) the implementation of two classifiers that work iteratively, each one exploiting the decision of the other to improve the training phase, and (2) the use of relational features based on the current labeling and on the spatial structure of the image. The two classifiers are fed with the spectral features and with the spatial features, respectively. The spatial features are built using the relative abundance of each class in a neighborhood of the pixel (homogeneity index), where the neighborhood is properly defined. An important contribution to the success of the method is the adoption of a multiclass classifier, the multinomial logistic regression, and a proper use of the posterior probabilities to infer the class labeling and build the relational data. The results of the two classifiers are eventually combined by means of an ensemble decision. The algorithm has been successfully tested on three standard hyperspectral images taken from the Airborne Visible-Infrared Imaging Spectrometer and ROSIS airborne sensors and compared with classification algorithms recently proposed in the literature.

Focusing algorithms analysis for Ground-Based SAR images

Ground-Based Synthetic Aperture Radar (GB-SAR) is emerging as a relatively new system for specific short ranges and frequent monitoring applications. The main GB-SAR focusing algorithms proposed in literature are compared in terms of focusing quality and ability to manage the approximation of the space-varying two-dimensional kernel. Residual migration and phase errors which have implications on the target impulse response deformation, are given both in analytic and numerical forms, using simulations and real data acquired with a Ku-band system.

Persistent point scatterer statistical analysis for X-band SAR data: the Cosmo-SkyMed case study

This article addresses the problem of characterizing the statistical properties of persistent point scatterers (PPSs) retrieved in a stack of interferometric synthetic aperture radar (SAR) images. The set of parameters are exploited to obtain information about the radiometric response to microwave excitation and to infer some properties related to the physical nature of the targets.The permanent scatterers (PSs) technique is a well-known and worthwhile tool in the field of SAR interferometry and is used to detect terrain deformation with millimetrical accuracy. Although PSs have been adopted in many applications, the physical nature of these targets is still a matter of investigation. A good knowledge of the target properties, such as shape, size, orientation, and roughness, can be a key step for the correct interpretation of the terrain deformation mechanisms and for object identification. PPSs emerge as a subset of PSs that are characterized by a more stringent property; they have an impulsive trend both along and across track directions and sufficient isolation from other strong scatterers. These targets deserve particular attention, since few of them are able, for example, to provide high-precision antenna pattern shape and pointing estimation as well as accurate terrain deformation along the temporal baseline. In this work, the PPS intensity, signal-to-noise ratio (SNR), amplitude dispersion index, and antenna similarity index (a newly defined parameter) are statistically characterized for different kinds of land cover. A stack of images from the Cosmo-SkyMed constellation of satellites are exploited to obtain results for the PPS statistics on four different area typologies: full urban, suburban, vegetated and sparse buildings, and mountains with possible forested areas.

Impact of focusing of ground-based SAR data on the quality of interferometric SAR applications

A Ground-Based Synthetic Aperture Radar (GB-SAR) is nowadays employed in several applications. The processing of ground-based, space and airborne SAR data relies on the same physical principles. Nevertheless specific algorithms for the focusing of data acquired by GB-SAR system have been proposed in literature. In this work the impact of the main focusing methods on the interferometric phase dispersion and on the coherence has been studied by employing a real dataset obtained by carrying out an experiment. Several acquisitions of a scene with a corner reflector mounted on a micrometric screw have been made; before some acquisitions the micrometric screw has been displaced of few millimetres in the Line-of-Sight direction. The images have been first focused by using two different algorithms and correspondently, two different sets of interferograms have been generated. The mean and standard deviation of the phase values in correspondence of the corner reflector have been compared to those obtained by knowing the real displacement of the micrometric screw. The mean phase and its dispersion and the coherence values for each focusing algorithm have been quantified and both the precision and the accuracy of the interferometic phase measurements obtained by using the two different focusing methods have been assessed.

Data driven analysis of functional brain networks in fMRI for schizophrenia investigation

The purpose of this article is to present a methodology to identify the sources of activity in brain networks from functional magnetic resonance imaging (fMRI) data using the multiset canonical correlation analysis algorithm. The aim is to lay the foundations for a screening marker to be used as indicator of mental diseases. Group analysis blind source separation methods have proved reliable to extract the latent sources underlying the brain activities but currently there is no recognized biomarker for mental disorders. Recent studies have identified alterations in the so called default mode network (DMN) that are common to several neuropsychiatric disorders, including schizophrenia. In particular, here we account for the hypothesis that the alterations in the DMN activity can be effectively highlighted by analyzing the transient states between two different tasks. A set of fMRI data acquired from 18 subjects performing working memory tasks is investigated for such purpose. Subjects are patients affected by schizophrenia for one half and healthy control subjects for the other. Under these conditions, the proposed methodology provides high discrimination performances in terms of classification error, thereby providing promising results for a preliminary tool able to monitor the disease state or to perform a prescreening for patients at risk for schizophrenia.

Use of advanced SAR monitoring techniques for the assessment of the behaviour of old embankment dams

The need of reliable monitoring of old embankment dams is rapidly increasing since a large number of these structures are still equipped with old monitoring devices, usually installed some decades ago, which are generally capable to provide only localized information on specific areas of the embankment. This work discusses the use of Ground-Based Synthetic Aperture Radar (GBSAR) interferometry technique to observe and control the structural behavior of earthfill or rockfill embankments for dam impoundments. This non-invasive technique provides displacements patterns measured with sub-millimeter precision. Monitoring strategies of earthfill dam embankment in Southern Italy are presented.

Functional brain networks and schizophrenia analysis with fMRI by Multiset Canonical Correlation Analysis

In this work the multivariate technique called Multiset Canonical Correlation Analysis (M-CCA) is applied to study a group of functional Magnetic Resonance Imaging (fMRI) datasets acquired during a set of working memory (WM) tasks. The examined subjects are a small group of schizophrenic patients and an equal number of controls (healthy subjects). The purpose of the paper is to show that M-CCA is able to identify specific areas of the brain network that can be related with the sources of activity during the WM task. It is also shown that the different degree of activation between the two groups allows to discriminate between Controls and Patients and then provides a promising way for a pre-screening classification of successive dataset. This result can contribute, in future, to populate a database of specific brain activation areas of mental disorders.

Focusing SAR images by compressive sensing: Study of interferometric properties

In this work we study the problem of focusing Synthetic Aperture Radar (SAR) images by means of Compressive Sensing (CS). The presented technique is introduced as an alternative to the traditional focusing methods, suggesting new modes for data acquisition in a more efficient configurations or on-board processing in case of spaceborne sensors. In this paper the method is tested on both simulated and real images acquired by a Ground Based Synthetic Aperture Radar (GB-SAR), for which images of reduced size can be generated with no difficulty. Results of comparison of CS processing with an exact focusing algorithm are shown in terms of root mean square error of amplitude and phase as a function of the number of focused targets and undersampling of the acquisition lines. Coherence of a CS-processed couple of images is also evaluated. The purpose of the paper is to show the potential of CS applied to SAR systems, regardless (at the moment) of the efficiency in computational load. In particular, we show that the image can be reconstructed without loss of resolution after dropping a fair percentage of the received pulses.

Optimised data structures for large scale content-based geo-indexing

Image mining consists of the procedures that allow to access, search and explore very large databases of data. Institutions like spatial agencies have to manage huge archives of Earth Observation (EO) images and need solutions to make data available to users from both the algorithmic and the infrastructural point of views. On the other side, users would need to explore the variety of images not just based on metadata, like time of acquisition or sensor parameters, but also by getting knowledge of their content. In this contribution, we investigate methodologies for content-based EO image retrieval via example-based queries. In particular, we present a procedure for the indexing of large-scale unstructured archives, built on top of a cluster analytics framework, Apache Spark. The procedure is based on a hierarchical and scalable implementation of a space partitioning algorithm and allows O(log n) response query times. Scalability analyses are conducted on polarimetric data from NASA/JPL archives, by using virtualized computing resources distributed over the Internet. In particular, the effects of the cluster size and of the hardware scale-up are demonstrated. The results also reveal the applicative potential of using on-demand cloud-based resources.

Automatic registration of multispectral images through maximization of mutual information

In this paper we propose a method to get fine registration of high resolution multispectral images. The algorithm supposes that a coarse registration, based on ancillary information, has been already performed. It is known, in fact, that residual distortions remain, due to the combined effects of Earth rotation and curvature, view geometry, sensor operation, variations in platform velocity, atmospheric and terrain effects. The algorithm grounds its main idea on the information-theoretic approach to register volumetric medical images of different modalities. Registration is achieved by adjustment of the relative position and orientation until the mutual information between the images is maximized. The idea is that the join information is maximized when the two images are at their best registration. This approach works directly with image data but in principle it can be applied in any transformed domain. While the original algorithm has been thought to make registration in a limited search space (i.e. translation and orientation), in the remote sensing framework the class of transformations is extended allowing scaling, shearing or a general polynomial model. The maximization of the target function is performed using both the stochastic gradient descent algorithm and the simulated annealing, since the former is known to occasionally deadlock in local maxima. We have applied the algorithm on a SPOT-5 couple of images, achieving the registration of chips of size 256x256 pixels at time. Accuracy has been obtained comparing the results with the outcomes of a commercial software that adopts a sort of Normalized Cross-Correlation method. On 143 chips taken throughout the image, the final translation accuracy resulted well below 1 pixel and the rotation accuracy about 0.015deg.