Tesi di Dottorato

Permanent URI for this communityTesi di Dottorato

Browse

Author: search.filters.author.Palopoli, Luigi

Search Results

Now showing 1 - 10 of 24
  • No Thumbnail Available
    Item
    Using multi-layer social networks for opportunistic routing
    (2012-10-24) Socievole, Annalisa; Palopoli, Luigi; Marano, Salvatore; De Rango, Floriano
  • No Thumbnail Available
    Item
    Designing Cloud services for data processing and knowledge discovery
    (2012-10-24) Marozzo, Fabrizio; Palopoli, Luigi; Talia, Domenico; Trunfio, Paolo
  • No Thumbnail Available
    Item
    Metodi di fusione dei dati per sistemi di assistenza alla guida
    (2012-10-24) Lupia, Marco; Palopoli, Luigi; Casavola, Alessandro
    Questo documento riporta tutta l'attivit a svolta durante il corso di dottora- to riguardo lo studio, l'applicazione e la sperimentazione di metodi di Data Fusion nell'ambito dei sistemi avanzati di assistenza alla guida per il miglio- ramento delle prestazioni e dell'a dabilit a. In linea di principio, sembra ra- gionevole a ermare che combinando in modo ottimale le informazioni proveni- enti da pi u sensori e possibile progettare e realizzare un sistema meno sensibile alle variazioni ambientali e a possibili errori di misura dovuti alla presenza di outliers. Tuttavia, dal punto di vista pratico i costi aggiuntivi che si devono sostenere, sia in termini di hardware necessario alla raccolta dei dati dai sen- sori addizionali che per la necessit a di disporre di sistemi di calcolo pi u potenti, potrebbero non essere giusti cati se i miglioramenti ottenuti nelle situazioni pi u probabili e pi u realistiche di utilizzo sono modesti. Il presente documento o re, innanzitutto, una panoramica su vari algoritmi di visione arti ciale utilizzati per il riconoscimento della segnaletica orizzon- tale e per la stima del tempo di invasione (tTLC). Quest'ultimo parametro gioca un ruolo determinante nell'avvertimento tempestivo del conducente in caso di superamento dei limiti della carreggiata. I vari algoritmi sono stati testati in varie condizioni di guida, valutandone le prestazioni conseguibili e il carico computazionale richiesto. Segue un'analisi dello stato dell'arte dei metodi e delle tecniche di Data Fu- sion pi u promettenti e che che meglio si prestano a migliorare l'accuratezza del calcolo della stima del tempo di invasione ttlc grazie alla disponibilit a di altri sensori oltre alla telecamera. Speci catamente, si sono confrontati i vari meto- di e algoritmi di Data Fusion, particolarizzati rispetto a vari modelli matem- atici della vettura e ai sensori disponibili, valutando le loro prestazioni in situ- azioni tipiche di guida e soprattutto rispetto all'errore percentuale di stima del tempo di invasione ttlc ottenuto, valutando anche il carico computazionale corrispondente. Gli algoritmi pi u promettenti sono stati implementati su piattaforma embedded.Questo documento riporta tutta l'attivit a svolta durante il corso di dottora- to riguardo lo studio, l'applicazione e la sperimentazione di metodi di Data Fusion nell'ambito dei sistemi avanzati di assistenza alla guida per il miglio- ramento delle prestazioni e dell'a dabilit a. In linea di principio, sembra ra- gionevole a ermare che combinando in modo ottimale le informazioni proveni- enti da pi u sensori e possibile progettare e realizzare un sistema meno sensibile alle variazioni ambientali e a possibili errori di misura dovuti alla presenza di outliers. Tuttavia, dal punto di vista pratico i costi aggiuntivi che si devono sostenere, sia in termini di hardware necessario alla raccolta dei dati dai sen- sori addizionali che per la necessit a di disporre di sistemi di calcolo pi u potenti, potrebbero non essere giusti cati se i miglioramenti ottenuti nelle situazioni pi u probabili e pi u realistiche di utilizzo sono modesti. Il presente documento o re, innanzitutto, una panoramica su vari algoritmi di visione arti ciale utilizzati per il riconoscimento della segnaletica orizzon- tale e per la stima del tempo di invasione (tTLC). Quest'ultimo parametro gioca un ruolo determinante nell'avvertimento tempestivo del conducente in caso di superamento dei limiti della carreggiata. I vari algoritmi sono stati testati in varie condizioni di guida, valutandone le prestazioni conseguibili e il carico computazionale richiesto. Segue un'analisi dello stato dell'arte dei metodi e delle tecniche di Data Fu- sion pi u promettenti e che che meglio si prestano a migliorare l'accuratezza del calcolo della stima del tempo di invasione ttlc grazie alla disponibilit a di altri sensori oltre alla telecamera. Speci catamente, si sono confrontati i vari meto- di e algoritmi di Data Fusion, particolarizzati rispetto a vari modelli matem- atici della vettura e ai sensori disponibili, valutando le loro prestazioni in situ- azioni tipiche di guida e soprattutto rispetto all'errore percentuale di stima del tempo di invasione ttlc ottenuto, valutando anche il carico computazionale corrispondente. Gli algoritmi pi u promettenti sono stati implementati su piattaforma embedded.
  • No Thumbnail Available
    Item
    Environmental and physiological parameters measurement in images and video
    (2012-10-24) Kurylyak, Yuriy; Palopoli, Luigi; Grimaldi, Domenico
    Measurement in images and video is a new challenging research direction. Up to now, cameras are mostly used as interaction devices. Computer vision technologies, however, can turn an ordinary video camera to a powerful tool for counting, measuring and inspecting. Using the camera as a measuring sensor is very interesting as allows creating a ”universal” measurement instrument, where new type of measurements can be added just by changing the software. Appearance of smartphones brings measurements in image and video to the new level, introducing a small, portable, autonomous measurement device. A lot of efforts have been made to convert smartphones to mobile tools for measuring the object length, width, size, angles, area, dimensions etc. This Ph.D. thesis investigates novel image and video processing techniques and shows how they can be used for non-invasive measurement of various environmental and physiological parameters. The three logical steps describe the possible types of measurements: in static image, in video and using smartphones. First, the case with a single image affected by a motion blur is considered and appropriate techniques for locating the regions with motion blur and parameters extraction are presented. A new method to detect the locally motion blurred regions from the image with complex still background is introduced. Analysis in the frequency domain, statistical analysis and windowing techniques are used to find blurred object, and the Fourier and Radon transformations are used to compute its motion characteristics. Analysis of video allows measuring additional characteristics of the objects that change over time. Monitoring of the human fatigue level is done by eyelid blinks detection and analysis. Two solutions are proposed: the non-invasive blink detection system based on infrared camera and webcam. The usage of infrared camera with switching light is used for fast and easy pupil detection in each frame, while the webcam is used to create a very cheap but still effective system. The problem of eyes detection is solved by using a cascade of boosted classifiers based on Haar-like features. The algorithm is proposed to detect closure and opening of the eyes and to distinguish voluntary blinks from the involuntary ones. Finally, the smartphone is used for photoplethysmogram acquisition and measurement of vital parameters. The proposed approach utilizes a concept of image acquisition similar to the one of a pulse oximeter. The problem of finger detection in video as well as verification of the proper usage of the system is solved by using colour segmentation in each colour channel. Then, the pulse rate is evaluated based on adaptive and statistical analysis. Moreover, the blood pressure is estimated by means of artificial neural network. A set of parameters are proposed to be extracted from the photoplethysmographic signal and used as the input of the neural network. For wide representation of training data the Multiparameter Intelligent Monitoring in Intensive Care waveform dataset is used.
  • No Thumbnail Available
    Item
    A spatial data infrastructure
    (2012-10-24) D'Amore, Francesco; Palopoli, Luigi; Talia, Domenico; Cinnirella, Sergio
  • No Thumbnail Available
    Item
    Metodi e algoritmi per il riconoscimento del livello di distrazione e affaticamento nella guida di autoveicoli
    (2012-10-24) Cario, Gianni; Palopoli, Luigi; Casavola, Alessandro
  • No Thumbnail Available
    Item
    New materials and technologies for compact antennas and circuits at millimeter frequencies
    (2012-10-24) Borgia, Antonio; Palopoli, Luigi; Costanzo, Sandra
  • No Thumbnail Available
    Item
    Probabilistic approaches to recommendations
    (2012-10-24) Barbieri, Nicola; Palopoli, Luigi; Greco, Sergio; Manco, Giuseppe
  • No Thumbnail Available
    Item
    Pattern extraction from data with application to image processing
    (2012-10-24) Amelio, Alessia; Palopoli, Luigi; Pizzuti, Clara
    The term Information Extraction refers to the automatic extraction of structured information from data. In such a context, the task of pattern extraction plays a key role, as it allows to identify particular trends and recurring structures of interest to a given user. For this reason, pattern extraction techniques are available in a wide range of applications, such as enterprise applications, personal information management, web oriented and scientific applications. In this thesis, analysis is focused on pattern extraction techniques from images and from political data. Patterns in image processing are defined as features derived from the subdivision of the image in regions or objects and several techniques have been introduced in the literature for extracting these kinds of features. Specifically, image segmentation approaches divide an image in ”uniform” region patterns and both boundary detection and region-clustering based algorithms have been adopted to solve this problem. A drawback of these methods is that the number of clusters must be predetermined. Furthermore, evolutionary techniques have been successfully applied to the problem of image segmentation. However, one of the main problems of such approaches is the determination of the number of regions, that cannot be changed during execution. Consequently, we formalize a new genetic graph-based image segmentation algorithm that, thanks to the new fitness function, a new concept of neighborhood of pixels and the genetic representation, is able to partition images without the need to set a priori the number of segments. On the other hand, some image compression algorithms, recently proposed in literature, extract image patterns for performing compression, such as extensions to 2D of the classical Lempel-Ziv parses, where repeated occurrences of a pattern are substituted by a pointer to that pattern. However, they require a preliminary linearization of the image and a consequent extraction of linear patterns. This could miss some 2D recurrent structures which are present inside the image. We propose here a new technique of image compression which extracts 2D motif patterns from the image in which also some pixels are omitted in order to increase the gain in compression and which uses these patterns to perform compression. About pattern extraction in political science, it consists in detecting voter profiles, ideological positions and political interactions from political data. Some proposed pattern extraction techniques analyze the Finnish Parliament and the United States Senate in order to discover political trends. Specifically, hierarchical clustering has been employed to discover meaningful groups of senators inside the United States Senate. Furthermore, different methods of community detection, based on the concept of modularity, have been used to detect the hierarchical and modular design of the networks of U.S. parliamentarians. In addition, SVD has been applied to analyze the votes of the U.S. House of Representatives. In this thesis, we analyze the Italian Parliament by using different tools coming from Data Mining and Network Analysis with the aim of characterizing the changes occurred inside the Parliament, without any prior knowledge about the ideology or political affiliation of its representatives, but considering only the votes cast by each parliamentarian.
  • No Thumbnail Available
    Item
    Distributed command governor strategies for multi-agent dynamical systems
    (2011-11-23) Francesco Tedesco, Francesco Tedesco; Palopoli, Luigi; Casavola, Alessandro
    This dissertation presents a class of novel distributed supervision strategies for multi-agent linear systems connected via data networks and subject to coordination constraints. Such a coordination-by-constraint paradigm is characterized by a set of spatially distributed dynamic systems, connected via communication channels, with possibly dynamical coupling amongst them which need to be supervised and coordinated in order to accomplish their overall objective. The basic design philosophy of the Command Governor (CG) set-point management is used here in order to maintain a pre-stabilized system within prescribed constraints. While in traditional CG schemes the set-point manipulation is undertaken on the basis of the actual measure of the state, in this dissertation it is shown that the CG design problem can be solved also in the case that such an explicit measure is not available by forcing the state evolutions to stay ”not too far” from the manifold of feasible steady-states. This approach, referred to as Feed-Forward CG (FF-CG), is a convenient solution to be used in distributed applications where the cost of measuring the overall state and distributing it amongst the agents may be a severe limitation. Several distributed strategies, based both on CG and FF-CG ideas, will be fully described and analyzed. First, we propose some “non-iterative” schemes in which the agents acting as supervisors communicate once during the decision process. In this respect, a “sequential” distributed strategy in which only one agent at the time is allowed to manipulate its own reference signal is proposed. Such a strategy, although interesting by itself in some applications, will be instrumental to introduce a more effective “parallel” distributed strategy, in which all agents are allowed, under certain conditions, to modify their own reference signals simultaneously. Then an “iterative” procedure, borrowed from the literature, has been here adapted in order to build more efficient distributed schemes which however require larger amount of data exchanges for their implementation. With the aim of evaluating the distributed methods here proposed, several cases of study involving the coordination autonomous vehicles, power networks and water networks management are illustrated.