Metodi di tipo level set per la segmentazione di immagini di nevi

La tesi in oggetto propone un algoritmo che viene incontro alla necessità di segmentare in modo regolare immagini di nevi. Si è fatto uso di metodi level set region-based in una formulazione variazionale. Tale metodo ha permesso di ottenere una segmentazione precisa, adattabile a immagini di nevi con caratteristiche molto diverse ed in tempi computazionali molto bassi.

Modeling and Model Predictive Control of an Industrial Hydraulic System

This thesis aims to illustrate the construction of a mathematical model of a hydraulic system, oriented to the design of a model predictive control (MPC) algorithm. The modeling procedure starts with the basic formulation of a piston-servovalve system. The latter is a complex non linear system with some unknown and not measurable effects that constitute a challenging problem for the modeling procedure. The first level of approximation for system parameters is obtained basing on datasheet informations, provided workbench tests and other data from the company. Then, to validate and refine the model, open-loop simulations have been made for data matching with the characteristics obtained from real acquisitions. The final developed set of ODEs captures all the main peculiarities of the system despite some characteristics due to highly varying and unknown hydraulic effects, like the unmodeled resistive elements of the pipes. After an accurate analysis, since the model presents many internal complexities, a simplified version is presented. The latter is used to linearize and discretize correctly the non linear model. Basing on that, a MPC algorithm for reference tracking with linear constraints is implemented. The results obtained show the potential of MPC in this kind of industrial applications, thus a high quality tracking performances while satisfying state and input constraints. The increased robustness and flexibility are evident with respect to the standard control techniques, such as PID controllers, adopted for these systems. The simulations for model validation and the controlled system have been carried out in a Python code environment.

Studio e sviluppo di un metodo di detezione della camera atriale in ecografia intracardiaca

Il lavoro di tesi si è svolto in collaborazione con il laboratorio di elettrofisiologia, Unità Operativa di Cardiologia, Dipartimento Cardiovascolare, dell’ospedale “S. Maria delle Croci” di Ravenna, Azienda Unità Sanitaria Locale della Romagna, ed ha come obiettivo lo sviluppo di un metodo per l’individuazione dell’atrio sinistro in sequenze di immagini ecografiche intracardiache acquisite durante procedure di ablazione cardiaca transcatetere per il trattamento della fibrillazione atriale. La localizzazione della parete posteriore dell'atrio sinistro in immagini ecocardiografiche intracardiache risulta fondamentale qualora si voglia monitorare la posizione dell'esofago rispetto alla parete stessa per ridurre il rischio di formazione della fistola atrio esofagea. Le immagini derivanti da ecografia intracardiaca sono state acquisite durante la procedura di ablazione cardiaca ed esportate direttamente dall’ecografo in formato Audio Video Interleave (AVI). L’estrazione dei singoli frames è stata eseguita implementando un apposito programma in Matlab, ottenendo così il set di dati su cui implementare il metodo di individuazione della parete atriale. A causa dell’eccessivo rumore presente in alcuni set di dati all’interno della camera atriale, sono stati sviluppati due differenti metodi per il tracciamento automatico del contorno della parete dell’atrio sinistro. Il primo, utilizzato per le immagini più “pulite”, si basa sull’utilizzo del modello Chan-Vese, un metodo di segmentazione level-set region-based, mentre il secondo, efficace in presenza di rumore, sfrutta il metodo di clustering K-means. Entrambi i metodi prevedono l’individuazione automatica dell’atrio, senza che il clinico fornisca informazioni in merito alla posizione dello stesso, e l’utilizzo di operatori morfologici per l’eliminazione di regioni spurie. I risultati così ottenuti sono stati valutati qualitativamente, sovrapponendo il contorno individuato all'immagine ecografica e valutando la bontà del tracciamento. Inoltre per due set di dati, segmentati con i due diversi metodi, è stata eseguita una valutazione quantitativa confrontatoli con il risultato del tracciamento manuale eseguito dal clinico.

L'informazione di fase per la segmentazione volumetrica di dati ecocardiografici

L’imaging ad ultrasuoni è una tecnica di indagine utilizzata comunemente per molte applicazioni diagnostiche e terapeutiche. La tecnica ha numerosi vantaggi: non è invasiva, fornisce immagini in tempo reale e l’equipaggiamento necessario è facilmente trasportabile. Le immagini ottenute con questa tecnica hanno tuttavia basso rapporto segnale rumore a causa del basso contrasto e del rumore caratteristico delle immagini ad ultrasuoni, detto speckle noise. Una corretta segmentazione delle strutture anatomiche nelle immagini ad ultrasuoni è di fondamentale importanza in molte applicazioni mediche . Nella pratica clinica l’identificazione delle strutture anatomiche è in molti casi ancora ottenuta tramite tracciamento manuale dei contorni. Questo processo richiede molto tempo e produce risultati scarsamente riproducibili e legati all’esperienza del clinico che effettua l’operazione. In ambito cardiaco l’indagine ecocardiografica è alla base dello studio della morfologia e della funzione del miocardio. I sistemi ecocardiografici in grado di acquisire in tempo reale un dato volumetrico, da pochi anni disponibili per le applicazioni cliniche, hanno dimostrato la loro superiorità rispetto all’ecocardiografia bidimensionale e vengono considerati dalla comunità medica e scientifica, la tecnica di acquisizione che nel futuro prossimo sostituirà la risonanza magnetica cardiaca. Al fine di sfruttare appieno l’informazione volumetrica contenuta in questi dati, negli ultimi anni sono stati sviluppati numerosi metodi di segmentazione automatici o semiautomatici tesi alla valutazione della volumetria del ventricolo sinistro. La presente tesi descrive il progetto, lo sviluppo e la validazione di un metodo di segmentazione ventricolare quasi automatico 3D, ottenuto integrando la teoria dei modelli level-set e la teoria del segnale monogenico. Questo approccio permette di superare i limiti dovuti alla scarsa qualità delle immagini grazie alla sostituzione dell’informazione di intensità con l’informazione di fase, che contiene tutta l’informazione strutturale del segnale.

Sea level measurement using single GNSS antenna SNR signals

This thesis presents a possible method to calculate sea level variation using geodetic-quality Global Navigate Satellite System (GNSS) receivers. Three antennas are used: two small antennas and a choke ring one, analyzing only Global Positioning System signals. The main goal of the thesis is to test a modified configuration for antenna set up. In particular, measurements obtained tilting one antenna to face the horizon are compared to measurements obtained from antennas looking upward. The location of the experiment is a coastal environment nearby the Onsala Space Observatory in Sweden. Sea level variations are obtained using periodogram analysis of the SNR signal and compared to synthetic gauge generated from two independent tide gauges. The choke ring antenna provides poor result, with an RMS around 6 cm and a correlation coefficients of 0.89. The smaller antennas provide correlation coefficients around 0.93. The antenna pointing upward present an RMS of 4.3 cm and the one pointing the horizon an RMS of 6.7 cm. Notable variation in the statistical parameters is found when modifying the length of the interval analyzed. In particular, doubts are risen on the reliability of certain scattered data. No relation is found between the accuracy of the method and weather conditions. Possible methods to enhance the available data are investigated, and correlation coefficient above 0.97 can be obtained with small antennas when sacrificing data points. Hence, the results provide evidence of the suitability of SNR signal analysis for sea level variation in coastal environment even in the case of adverse weather conditions. In particular, tilted configurations provides comparable result with upward looking geodetic antennas. A SNR signal simulator is also tested to investigate its performance and usability. Various configuration are analyzed in combination with the periodogram procedure used to calculate the height of reflectors. Consistency between the data calculated and those received is found, and the overall accuracy of the height calculation program is found to be around 5 mm for input height below 5 m. The procedure is thus found to be suitable to analyze the data provided by the GNSS antennas at Onsala.

Sea level trends in the Mediterranean from tide gauges and satellite altimetry

Sea level variation is one of the parameters directly related to climate change. Monitoring sea level rise is an important scientific issue since many populated areas of the world and megacities are located in low-lying regions. At present, sea level is measured by means of two techniques: the tide gauges and the satellite radar altimetry. Tide gauges measure sea-level relatively to a ground benchmark, hence, their measurements are directly affected by vertical ground motions. Satellite radar altimetry measures sea-level relative to a geocentric reference and are not affected by vertical land motions. In this study, the linear relative sea level trends of 35 tide gauge stations distributed across the Mediterranean Sea have been computed over the period 1993-2014. In order to extract the real sea-level variation, the vertical land motion has been estimated using the observations of available GPS stations and removed from the tide gauges records. These GPS-corrected trends have then been compared with satellite altimetry measurements over the same time interval (AVISO data set). A further comparison has been performed, over the period 1993-2013, using the CCI satellite altimetry data set which has been generated using an updated modeling. The absolute sea level trends obtained from satellite altimetry and GPS-corrected tide gauge data are mostly consistent, meaning that GPS data have provided reliable corrections for most of the sites. The trend values range between +2.5 and +4 mm/yr almost everywhere in the Mediterranean area, the largest trends were found in the Northern Adriatic Sea and in the Aegean. These results are in agreement with estimates of the global mean sea level rise over the last two decades. Where GPS data were not available, information on the vertical land motion deduced from the differences between absolute and relative trends are in agreement with the results of other studies.

New assessments of global sea level rise from tide gauges

Estimates of global sea-level change rates based on observations from Tide Gauges (TGs) show a long-term global mean sea-level rise (GMSLR) of 1÷2 mm/yr for the 20th century. The considerable scatter in these estimates is mainly attributable to the uneven distribution of the TG sites and to several physical phenomena that cause local sea level to deviate from the global mean, or to affect the TG record through land subsidence or uplift. The main cause of vertical ground motion on a regional space scale is the response of the Earth to past ice loads, called Glacial Isostatic Adjustment (GIA), which is often modelled and corrected for. In this work, a simple average approach was used to revisit two past estimates based on small sets of long, high-quality TG records in view of the longer record available, employing a newer GIA model (ICE-6G) from Peltier et al. [2015]. The value of GMSLR obtained from both sets is (1.5±0.4) mm/yr. In addition, a much larger set of TGs was used to estimate the contemporary (post 1993) GMSLR using satellite estimates from Cazenave et al. [2018] as a benchmark, in an attempt to understand how a simple average approach could perform for larger sets. The resulting estimate of (3.4÷3.5)±0.2 mm/yr (depending on the GIA correction applied) is comparable to the satellite result of (3.1±0.3) mm/yr.

Machine Learning algorithm on seat comfort level measurement, mapping pressure values with body regions

The comfort level of the seat has a major effect on the usage of a vehicle; thus, car manufacturers have been working on elevating car seat comfort as much as possible. However, still, the testing and evaluation of comfort are done using exhaustive trial and error testing and evaluation of data. In this thesis, we resort to machine learning and Artificial Neural Networks (ANN) to develop a fully automated approach. Even though this approach has its advantages in minimizing time and using a large set of data, it takes away the degree of freedom of the engineer on making decisions. The focus of this study is on filling the gap in a two-step comfort level evaluation which used pressure mapping with body regions to evaluate the average pressure supported by specific body parts and the Self-Assessment Exam (SAE) questions on evaluation of the person’s interest. This study has created a machine learning algorithm that works on giving a degree of freedom to the engineer in making a decision when mapping pressure values with body regions using ANN. The mapping is done with 92% accuracy and with the help of a Graphical User Interface (GUI) that facilitates the process during the testing time of comfort level evaluation of the car seat, which decreases the duration of the test analysis from days to hours.