Feature extraction from a tensor based local image representation introduced by Knutsson in [37] is discussed. The tensor representation keeps statements of structure, certainty of statement and energy separate. Further processing for obtaining new features also having these three entities separate is achieved by the use of a new concept, tensor field filtering. Tensor filters for smoothing and for extraction of circular symmetries are presented and discussed in particular. These methods are used for corner detection and extraction of more global features such as lines in images. A novel method for grouping local orientation estimates into global line parameters is introduced. The method is based on a new parameter space, the Möbius Strip parameter space, which has similarities to the Hough transform. A local centroid clustering algorithm is used for classification in this space. The procedure automatically divides curves into line segments with appropriate lengths depending on the curvature. A linked list structure is built up for storing data in an efficient way. / <p>Ogiltigt nummer / annan version: I publ. nr 290:s ISBN: 91-7870-815-X.</p>

No description available.

The ladle metallurgical furnace (LMF) is a very flexible and common unit operation found in most steelmaking melt shops, and enables the adjustment and fine-tuning of molten steel's composition and temperature prior to casting. Despite the importance of ladle metallurgy to the overall steel making process very little has been achieved in the way of advanced ladle control. Limited sensors are available to monitor heat progress during refining and current control methods involve manual procedures. This thesis represents part of an ongoing study on the modelling of a full-scale LMF in real-time with the forward goal of improved control and optimization. The first part of this thesis details a vision-based sensor for analyzing ladle eye dynamics online using a multivariate image analysis (MIA) technique based on principal component analysis (PCA). Predictive capabilities of the developed model are demonstrated using previously published cold model data over a wide range of operating variables. Further, preliminary work has confirmed the ability of the sensor for potential use in an industrial setting. The second part of this study concerns the development of metallurgical models for assessing the state of a ladle metallurgical furnace. Specifically, a multi-component kinetic model in combination with developed slag and steel thermodynamic solution models were used to quantitatively describe the kinetics of slag-metal reactions within 41 industrially sampled heats at ArcelorMittal Dofasco's LMF#2. Metal phase mass transfer coefficients for all elements in steel were assumed to follow the empirical relation derived from measured sulphur contents, while slag phase mass transfer coefficients were calculated by fitting the ratio of k^Mm/ k^MxOysl to the experimental results. On the basis of the fitted results, slag phase mass transfer coefficient correlations were evaluated using linear regression. Computed results from the model using these slag phase mass transfer coefficient correlations were found to be consistent with the experimental data. In regard to the developed thermodynamic solution models, original contributions to the modified interaction parameter formalism and cell model are presented. As process model predictions are invariably uncertain, the final part of this work involves the use of a stochastic model (extended Kalman filter) to account for process disturbances, model-mismatch and other sources of uncertainty that may result in significant error propagation causing poor process control and plant economics. Several case studies were performed to illustrate the effectiveness of the extended Kalman filter and its application to optimal sensor selection was introduced. / Thesis / Doctor of Philosophy (PhD)

Flexographic printing involves driving a paper substrate through the gap between two cylinders, where one of the cylinders has a print form layer covered in ink. The ink transfer from the print form to the paper is dependent on the pressure between the paper substrate and the cylinders. Knowledge about how the paper substrate reacts to compression is therefore important when aiming for high quality flexography printing. In this project the paper substrate in focus is corrugated board consisting of a top liner, a bottom liner and a fluted paper material in between the liners. To gain knowledge about the compression dynamics of the boards they are subjected to compression experiments with and without print form. The compression experiments are filmed and the videos are then analyzed using thresholding and tracking methods. The developed thresholding method is able to measure the thickness of each material separately in each frame while the developed tracking method is able to track the movement of the top and bottom metal blocks compressing the materials. The strains calculated from the image analysis measurements are then used as input values to the Maxwell spring-dashpot model with the intention of modelling the stress. Analysis of the compression experiment videos shows that the material properties of thecorrugated board may not be constant for the duration of the compression. This is because the fluting changes its shape when being compressed. Comparing the modelled stress to the stress measured by the press used in the experiments shows that the Maxwell spring-dashpot model is not able to model the stress accurately when the corrugated board goes through multiple stages of compression. The model works better for compression where the shape of the fluting is not significantly changed.

Fibrous air filters are commonly used to capture airborne particles due to their potential for a relatively high capture efficiency and low airflow resistance. Their performance characteristics make them ideal candidates in many instances, spanning a wide range from residential to sensitive industrial applications. However, as more particles are captured, the performance of the filter will evolve. This evolution of performance typically manifests as a higher capture efficiency and higher airflow resistance resulting from the additional particulate deposits. The prediction of fibrous filter performance has been the focus of research for many decades, resulting in numerous analytical, numerical, and empirical models. This work seeks to improve upon the state of aerosol filtration by investigating the process through which these models are developed and validated. To meet this objective, three major efforts are implemented: 1) a comprehensive literature review, 2) an aerosol and media measurement analysis focusing on instrumentation and Scanning Electron Microscope (SEM) imagery, and 3) the creation of a process to analyze and develop fibrous air filter models. A conceptual foundation is provided by the literature review, establishing the current state of fibrous filtration modeling of solid particles and identifying candidate models for implementation. The influence of data collection and reduction methodology for particle mass loading experiments is explored with an emphasis on the resulting effects towards filtration model development. Furthermore, an automated methodology to measure the physical characteristics of high efficiency particulate air (HEPA) filtration media is investigated, completing the set of variables necessary to predict filtration performance. Finally, an algorithm is proposed to optimize and correlate model variables to collected empirical data, allowing for the improvement of model predictions by investigating model functionality and identifying limitations. Altogether, the three efforts provide a framework through which fibrous aerosol filtration models of solid particles may be developed, validated, and systematically analyzed.

Seit der Erfindung des Mikroskops haben mikroskopische Bilder zu neuen Erkenntnissen in der biomedizinischen Forschung geführt. Moderne Mikroskope sind in der Lage große Bilddatensätze von zunehmender Komplexität zu erzeugen, was eine manuelle Analyse ineffizient, wenn nicht gar undurchführbar macht. In dieser Arbeit stelle ich zwei neue Methoden für die automatische Bildanalyse von Mikroskopiedaten vor. 1. Die Fourier-Ringkorrelations-basierte Qualitätsschätzung (FRC-QE), ist eine neue Metrik für die automatisierte Bildqualitätsschätzung von 3D-Fluoreszenzmikroskopieaufnahmen, hier getestet am Beispiel von menschlichen Hirnorganoiden. FRC-QE automatisiert die Qualitätskontrolle, eine Aufgabe, die häufig manuell durchgeführt wird und somit einen Engpass bei der Skalierung bildbasierter Experimente auf tausend oder mehr Bilder darstellt. Die Methode kann die Clearing-Effizienz über experimentelle Replikate und Protokolle quantifizieren. Sie ist auf verschiedene Mikroskopiemodelle übertragbar und lässt sich effizient auf Tausende von Bildern skalieren. 2. Der von mir entwickelte "WormObserver" ermöglicht Langzeitaufnahmen, verarbeitet automatisch die aufgenommenen Videos und erleichtert die Datenintegration über Tausende von Individuen hinweg, um Verhaltensmuster zu entschlüsseln. Darauf aufbauend, habe ich mich auf ein Beispiel für die Plastizität des Nervensystems konzentriert: Die Verhaltenstrajektorie des "C. elegans Dauer Exits". Um den Entscheidungsmechanismus beim Verlassen des Dauer Larvenstadiums zu charakterisieren, habe ich Zeitrafferdaten von Larvenpopulationen in verschiedenen Umgebungen erfasst, analysiert und wichtige Entscheidungspunkte identifiziert. Indem ich die Verhaltensanpassung mit der Genexpression kontextualisiert habe, konnte ich neue Erkenntnisse gewinnen, wie ein sich entwickelndes Nervensystem externe Stimuli robust integrieren und das Verhalten des Organismus an neue Umgebungen anpassen kann. / Since the invention of the microscope, microscopy images have generated new insights in biomedical research. While in the past these images were used for illustrative purposes, state-of-the-art microscopy images provide quantitative measurements. Moreover, modern microscopes are capable of autonomously producing large image datasets of increasing complexity, rendering manual analysis inefficient if not infeasible. Thus, extracting biologically relevant information from these datasets requires computational analysis using appropriate algorithms and software. While some analysis methods generalize to different microscope set-ups and types of images, others need to be well tailored to a particular problem. In this work, I present two new methods for automated image analysis of microscopy data. First, Fourier ring correlation-based quality estimation (FRC-QE) is a new metric for automated image quality estimation of 3D fluorescence microscopy acquisitions. I benchmarked the method in the context of evaluating clearing efficiency in human brain organoids. FRC-QE automates image quality control, a task that is often performed manually and thereby represents a bottleneck when scaling image-based experiments to thousand or more images. The method can estimate clearing efficiency across experimental replicates and clearing protocols. It generalizes to different microscopy modalities and efficiently scales to thousands of images. Second, I have developed a new method for behavioral imaging of C. elegans larvae. The “WormObserver” enables long-term imaging (>12h, >80k images/experiment), automatically processes the acquired videos a

Single-cell bio-analytical techniques play a pivotal role in contemporary biological and biomedical research. Among current high-throughput single-cell imaging methods, coherent Raman imaging offers both high bio-compatibility and high-throughput information-rich capabilities, offering insights into cellular composition, dynamics, and function. Coherent Raman imaging finds its value in diverse applications, ranging from live cell dynamic imaging, high-throughput drug screening, fast antimicrobial susceptibility testing, etc. In this thesis, I first present a deep learning algorithm to solve the inverse problem of getting a chemically labeled image from a single-shot femtosecond stimulated Raman scattering (SRS) image. This method allows high-speed, high-throughput tracking of lipid droplet dynamics and drug response in live cells. Second, I provide image-based single-cell analysis in an engineered Escherichia coli (E. coli) population, confirming the chemical composition and subcellular structure organization of individual engineered E. coli cells. Additionally, I unveil metabolon formation in engineered E. coli by high-speed spectroscopic SRS and two-photon fluorescence imaging. Lastly, I present stimulated Raman-activated cell ejection (S-RACE) by integrating high-throughput SRS imaging, in situ image decomposition, and high-precision laser-induced cell ejection. I demonstrate the automatic imaging-identification-sorting workflow in S-RACE and advance its compatibility with versatile samples ranging from polymer particles, single live bacteria/fungus, and tissue sections. Collectively, these efforts demonstrate the valuable capability of SRS in high-throughput single-cell imaging and sorting, opening opportunities for a wide range of biological and biomedical applications.

Note:

No description available.

No description available.