En exploitant la fonction d’amplification intrinsèque fournie par les transistors UTBB FDSOI, nous avons présenté des ISFET ultra sensibles. L'intégration de la fonctionnalité de détection a été réalisée en back end of line (BEOL), ce qui offre les avantages d'une fiabilité et d'une durée de vie accrues du capteur, d'une compatibilité avec le processus CMOS standard et d'une possibilité d'intégration d'un circuit diviseur capacitif. Le fonctionnement des MOSFETs, sans une polarisation appropriée de la grille avant, les rend vulnérables aux effets de grilles flottantes indésirables. Le circuit diviseur capacitif résout ce problème en polarisant la grille avant tout en maintenant la fonctionnalité de détection sur la même grille par un couplage capacitif au métal commun du BEOL. Par conséquent, le potentiel au niveau du métal BEOL est une somme pondérée du potentiel de surface au niveau de la grille de détection et de la polarisation appliquée au niveau de la grille de contrôle. Le capteur proposé est modélisé et simulé à l'aide de TCAD-Sentaurus. Un modèle mathématique complet a été développé. Il fournit la réponse du capteur en fonction du pH de la solution (entrée du capteur) et des paramètres de conception du circuit diviseur capacitif et du transistor UTBB FDSOI. Dans ce cas, des résultats cohérents ont été obtenus des travaux de modélisation et de simulation, avec une sensibilité attendue de 780 mV / pH correspondant à un film de détection ayant une réponse de Nernst. La modélisation et la simulation du capteur proposé ont également été validées par une fabrication et une caractérisation du capteur de pH à grille étendue avec validation de son concept. Ces capteurs ont été développés par un traitement séparé du composant de détection de pH, qui est connecté électriquement au transistor uniquement lors de la caractérisation du capteur. Ceci permet une réalisation plus rapide et plus simple du capteur sans avoir besoin de masques et de motifs par lithographie. Les capteurs à grille étendue ont présenté une sensibilité de 475 mV/pH, ce qui est supérieur aux ISFET de faible puissance de l'état de l’art. Enfin, l’intégration de la fonctionnalité de détection directement dans le BEOL des dispositifs FDSOI UTBB a été poursuivie. Une sensibilité expérimentale de 730 mV/pH a été obtenue, ce qui confirme le modèle mathématique et la réponse simulée. Cette valeur est 12 fois supérieure à la limite de Nernst et supérieure aux capteurs de l'état de l’art. Les capteurs sont également évalués pour la stabilité, la résolution, l'hystérésis et la dérive dans lesquels d'excellentes performances sont démontrées. / Exploiting the intrinsic amplification feature provided by UTBB FDSOI transistors, we demonstrated ultrahigh sensitive ISFETs. Integration of the sensing functionality was made in the BEOL which gives the benefits of increased reliability and life time of the sensor, compatibility with the standard CMOS process, and possibility for embedding a capacitive divider circuit. Operation of the MOSFETs without a proper front gate bias makes them vulnerable for undesired floating body effects. The capacitive divider circuit addresses these issues by biasing the front gate simultaneously with the sensing functionality at the same gate through capacitive coupling to a common BEOL metal. Therefore, the potential at the BEOL metal would be a weighted sum of the surface potential at the sensing gate and the applied bias at the control gate. The proposed sensor is modeled and simulated using TCAD-Sentaurus. A complete mathematical model is developed which provides the output of the sensor as a function of the solution pH (input to the sensor), and the design parameters of the capacitive divider circuit and the UTBB FDSOI transistor. In that case, consistent results have been obtained from the modeling and simulation works, with an expected sensitivity of 780 mV/pH corresponding to a sensing film having Nernst response. The modeling and simulation of the proposed sensor was further validated by a proof of concept extended gate pH sensor fabrication and characterization. These sensors were developed by a separated processing of just the pH sensing component, which is electrically connected to the transistor only during characterization of the sensor. This provides faster and simpler realization of the sensor without the need for masks and patterning by lithography. The extended gate sensors showed 475 mV/pH sensitivity which is superior to state of the art low power ISFETs. Finally, integration of the sensing functionality directly in the BEOL of the UTBB FDSOI devices was pursued. An experimental sensitivity of 730 mV/pH is obtained which is consistent with the mathematical model and the simulated response. This is more than 12-times higher than the Nernst limit, and superior to state of the art sensors. Sensors are also evaluated for stability, resolution, hysteresis, and drift in which excellent performances are demonstrated.

Structural health monitoring and building assessment are crucial to acquire structures’ states and maintain their conditions. Besides human-labor surveys that are subjective, time-consuming, and expensive, autonomous image and video analysis is a faster, more efficient, and non-destructive way. This thesis focuses on crack detection from videos, crack segmentation from images, and building assessment from street view images. For crack detection from videos, three approaches are proposed based on local binary pattern (LBP) and support vector machine (SVM), deep convolution neural network (DCNN), and fully-connected network (FCN). A parametric Naïve Bayes data fusion scheme is introduced that registers video frames in a spatiotemporal coordinate system and fuses information based on Bayesian probability to increase detection precision. For crack segmentation from images, the rotation-invariant property of crack is utilized to enhance the segmentation accuracy. The architectures of several approximately rotation-invariant DCNNs are discussed and compared using several crack datasets. For building assessment from street view images, a framework of multiple DCNNs is proposed to detect buildings and predict their attributes that are crucial for flood risk estimation, including founding heights, foundation types (pier, slab, mobile home, or others), building types (commercial, residential, or mobile home), and building stories. A feature fusion scheme is proposed that combines image feature with meta information to improve the predictions, and a task relation encoding network (TREncNet) is introduced that encodes task relations as network connections to enhance multi-task learning.

We use a Fully Convolutional Neural Network (FCNN) to classify pixels in historical document images, enabling the extraction of high-quality, pixel-precise and semantically consistent layers of masked content. We also analyze a dataset of hand-labeled historical form images of unprecedented detail and complexity. The semantic categories we consider in this new dataset include handwriting, machine-printed text, dotted and solid lines, and stamps. Segmentation of document images into distinct layers allows handwriting, machine print, and other content to be processed and recognized discriminatively, and therefore more intelligently than might be possible with content-unaware methods. We show that an efficient FCNN with relatively few parameters can accurately segment documents having similar textural content when trained on a single representative pixel-labeled document image, even when layouts differ significantly. In contrast to the overwhelming majority of existing semantic segmentation approaches, we allow multiple labels to be predicted per pixel location, which allows for direct prediction and reconstruction of overlapped content. We perform an analysis of prevalent pixel-wise performance measures, and show that several popular performance measures can be manipulated adversarially, yielding arbitrarily high measures based on the type of bias used to generate the ground-truth. We propose a solution to the gaming problem by comparing absolute performance to an estimated human level of performance. We also present results on a recent international competition requiring the automatic annotation of billions of pixels, in which our method took first place.

We use a Fully Convolutional Neural Network (FCNN) to classify pixels in historical document images, enabling the extraction of high-quality, pixel-precise and semantically consistent layers of masked content. We also analyze a dataset of hand-labeled historical form images of unprecedented detail and complexity. The semantic categories we consider in this new dataset include handwriting, machine-printed text, dotted and solid lines, and stamps. Segmentation of document images into distinct layers allows handwriting, machine print, and other content to be processed and recognized discriminatively, and therefore more intelligently than might be possible with content-unaware methods. We show that an efficient FCNN with relatively few parameters can accurately segment documents having similar textural content when trained on a single representative pixel-labeled document image, even when layouts differ significantly. In contrast to the overwhelming majority of existing semantic segmentation approaches, we allow multiple labels to be predicted per pixel location, which allows for direct prediction and reconstruction of overlapped content. We perform an analysis of prevalent pixel-wise performance measures, and show that several popular performance measures can be manipulated adversarially, yielding arbitrarily high measures based on the type of bias used to generate the ground-truth. We propose a solution to the gaming problem by comparing absolute performance to an estimated human level of performance. We also present results on a recent international competition requiring the automatic annotation of billions of pixels, in which our method took first place.

This master's thesis is focused on fully automatic calibration of traffic surveillance camera, which is used for speed measurement of passing vehicles. Thesis contains and describes theoretical information and algorithms related to this issue. Based on this information and algorithms, a comprehensive system design for automatic calibration and speed measurement was built. The proposed system has been successfully implemented. The implemented system is optimized to process the smallest portion of the video input for the automatic calibration of the camera. Calibration parameters are obtained after processing only two and half minutes of input video. The accuracy of the implemented system was evaluated on the dataset BrnoCompSpeed. The speed measurement error using the automatic calibration system is 8.15 km/h. The error is mainly caused by inaccurate scale acquisition, and when it is replaced by manually obtained scale, the error is reduced to 2.45 km/h. The speed measuring system itself has an error of only 1.62 km/h (evaluated using manual calibration parameters).

In this work is designed programme unit for acquisition analog acoustic signlas from sensors. There is used accurate A/D converter ADS1287 by the company Texas Iinstrument with resolution of 24 bits to digitizing these signals. There is used 32-bit microprocessor AT91SAM7S64 by the company Atmel to control this A/D converter and sending digitized data to PC. This microprocessor has implemented USB interface. By force of developmental programme units whit microprocessor and A/D converter is produced programme for microprocessor in developmental setting IAR Embedded Workbench IDE 5.0 and simple aplication for PC in setting Borland C++ Builder. Both of those programs are in language C++. Rate of sampling is 26 kHz. On the basis is realized programme unit USB-ADC whit microprocessor and A/D converter.

The aim of this thesis is to design a universal analog oscillator using transconductance amplifiers. For studying behaviour of chaotic dynamical systems can be used systems Class C. Suitable way for the purpose modeling dynamic phenomena arising in these systems is an electronic circuit that exhibits the same behavior as modeled system. After familiarization with the basic principles of synthesis of integrators systems, and studying the involvement of frequently used functional blocks were designed the concept of universal chaotic oscillator using transconductance amplifiers. The functionality of this circuit has been verified by PSpice simulation program. A typical feature of chaotic oscillator is extremely sensitivity to initial conditions. Each small change on the initial parameters can lead to major change in the shape of the attractor. The result of this thesis is a functional sample of a universal chaotic oscillator, which was verified by the dynamic behavior of the given differential equations.

The master thesis deals with designs of the single-ended and the fully-differential secondorder frequency filters with adjustable parameters. This filters operate in the current mode where main element is the current amplifier DACA (Digitally Adjustable Current Amplifier). The first part describes basic behaviour of frequency filters and distribution according to filter funcion, used parts and structure. The second part of thesis describes the design of the frequency filters by the M-C signal-flow graphs and the transformating single-ended structures to the fully-differential forms. The next part deals with properties of the active elements and their simulation models. The following part describe six circuits of the frequency filters with adjustable parameters and their simulations. Four simulation models with different properties were used for simulating of the DACA in the each circuits. The last part deals with the practical design and measurement of the selected frequency filters.

This master’s thesis deals with designing of fully-differential current-mode frequency filters using signal-flow graphs method. The first part is focused on a general description of frequency filters, its function and division. Active elements that create these frequency filters, such as multi-output current follower (MO-CF), balanced or multi-output transconductance amplifier (BOTA, MOTA) and digitally adjustace current amplifier (DACA) are described in the following part. Next, problems and various techniques of designing such filters are discussed on a theoretic basis. In the remaining part of the thesis there are six circuits of frequency filters described in detail; these connections are then transferred of passive elements to a proposed connections, with the assistance of a so-called reflection. Calculations of passive form are stated, as well as results of simulations, where nondifferential and differential variations of these designed frequency filters are compared. Finaly, it has been selected some variants of designs, which has been manufactured, then measured out and resulet has been compare between each other.

This thesis deals with the assessment of reinforced concrete elements using a fully probabilistic approach and its comparison with the method of partial reliability factors. This method of ultimate limit state design will be applied to determine the ultimate limit state of reinforced concrete column loaded by a combination of "N + M", the normal force and bending moment. Specifically, I will solve a circular reinforced concrete column with and without the effect of wrapping unidirectional carbon fiber fabrics and comparing the theoretically determined values with the values determined by experiment, which is carried out at the Institute of Concrete and Masonry Structures Faculty of Civil Engineering of the Technical University in Brno.