In this work, sensor tasking refers to assigning the times and pointing directions for a sensor to collect observations of cataloged objects, in order to maintain the accuracy of the orbit estimates. Sensor tasking must consider the dynamics of the objects and uncertainty in their positions, the coordinate frame in which the sensor tasking is defined, the timing requirements for observations, the sensor capabilities, the local visibility, and constraints on the information processing and communication. This research focuses on finding efficient ways to solve the sensor tasking optimization problem. First, different coordinate frames are investigated, and it is shown that the observer fixed Local Meridian Equatorial (ground-based) and Satellite Meridian Equatorial (space-based) coordinate frames provide consistent sets of pointing directions and accurate representations of orbit uncertainty for use by the optimizers in solving the sensor tasking problem. Next, two classical optimizers (greedy and Weapon-Target Assignment) which rely on convexity are compared with two Machine Learning optimizers (Ant Colony Optimization and Distributed Q-learning) which attempt to learn about the solution space in order to approximate a global optimal solution. It is shown that the learning optimizers are able to generate better solutions, while the classical optimizers are more efficient to run and require less tuning to implement. Finally, the realistic scenario where the optimization algorithm receives no feedback before it must make the next decision is introduced. The Predicted Measurement Probability (PMP) is developed, and employed in a two sensor optimization framework. The PMP is shown to provide effective feedback to the optimization algorithm regarding the observations of each sensor.<br>

The ability to numerically determine the received signal in an ultrasound pulse-echo system is very important for the development of new ultrasound applications, such as tissue characterization, complex object recognition, and identification of surface topology. The output signal from an ultrasound pulse-echo system depends on the transducer geometry, reflector shape, location and orientation, among others, therefore, only by numerical modeling can the output signal for a given measurement configuration be predicted. This thesis concerns about the numerical modeling and optimal design of annular array based ultrasound pulse-echo system for object recognition. Two numerical modeling methods have been implemented and evaluated for calculating received signal in a pulse-echo system. One is the simple, but computationally demanding Huygens Method and the other one is the computationally more efficient Diffraction Response for Extended Area Method (DREAM). The modeling concept is further extended for pulse-echo system with planar annular array. The optimal design of the ultrasound pulse-echo system is based on annular array transducer that gives us the flexibility to create a wide variety of insonifying fields and receiver characteristics. As the first step towards solving the optimization problem for general conditions, the problem of optimally identifying two specific reflectors is investigated. Two optimization methods, the straightforward, but computationally intensive Global Search Method and the efficient Waveform Alignment Method, have been investigated and compared.

<p>Although consciousness has been studied since ancient time, how the brain implements consciousness is still considered a great mystery by most. This thesis investigates the neural correlates of consciousness by measuring brain activity with functional magnetic resonance imaging (fMRI) while specific contents of consciousness are defined and maintained in various experimental settings. Study 1 showed that the brain works differently when creating a new conscious percept compared to when maintaining the same percept over time. Specifically, sensory and fronto-parietal regions were activated for both conditions but with different activation patterns within these regions. This distinction between creating and maintaining a conscious percept was further supported by Study 2, which in addition showed that there are both differences and similarities in how the brain works when defining a visual compared to an auditory percept. In particular, frontal cortex was commonly activated while posterior cortical activity was modality specific. Study 3 showed that task difficulty influenced the degree of frontal and parietal cortex involvement, such that fronto-parietal activity decreased as a function of ease of identification. This is interpreted as evidence of the non-necessity of these regions for conscious perception in situations where the stimuli are distinct and apparent. Based on these results a model is proposed where sensory regions interact with controlling regions to enable conscious perception. The amount and type of required interaction depend on stimuli and task characteristics, to the extent that higher-order cortical involvement may not be required at all for easily recognizable stimuli.</p>

Although consciousness has been studied since ancient time, how the brain implements consciousness is still considered a great mystery by most. This thesis investigates the neural correlates of consciousness by measuring brain activity with functional magnetic resonance imaging (fMRI) while specific contents of consciousness are defined and maintained in various experimental settings. Study 1 showed that the brain works differently when creating a new conscious percept compared to when maintaining the same percept over time. Specifically, sensory and fronto-parietal regions were activated for both conditions but with different activation patterns within these regions. This distinction between creating and maintaining a conscious percept was further supported by Study 2, which in addition showed that there are both differences and similarities in how the brain works when defining a visual compared to an auditory percept. In particular, frontal cortex was commonly activated while posterior cortical activity was modality specific. Study 3 showed that task difficulty influenced the degree of frontal and parietal cortex involvement, such that fronto-parietal activity decreased as a function of ease of identification. This is interpreted as evidence of the non-necessity of these regions for conscious perception in situations where the stimuli are distinct and apparent. Based on these results a model is proposed where sensory regions interact with controlling regions to enable conscious perception. The amount and type of required interaction depend on stimuli and task characteristics, to the extent that higher-order cortical involvement may not be required at all for easily recognizable stimuli.

Invariant object recognition is maybe the most basic and fundamental property of our visual system. It is the basis of many other cognitive tasks, like motor actions and social interactions. Hence, the theoretical understanding and modeling of invariant object recognition is one of the central problems in computational neuroscience. Indeed, object recognition consists of two different tasks: classification and identification. The focus of this thesis is on object identification under the basic geometrical transformations shift, scaling, and rotation. The visual system can perform shift, size, and rotation invariant object identification. This thesis consists of two parts. In the first part, we present and investigate the VisNet model proposed by Rolls. The generalization problems of VisNet triggered our development of a new neural network model for invariant object identification. Starting point for an improved generalization behavior is the search for an operation that extracts images features that are invariant under shifts, rotations, and scalings. Extracting invariant features guarantees that an object seen once in a specific pose can be identified in any pose. We present and investigate our model in the second part of this thesis.

Die automatisierte Objektidentifikation stellt ein modernes Werkzeug in den Geoinformationswissenschaften dar (BLASCHKE et al., 2012). Um bei thematischen Kartierungen untereinander vergleichbare Ergebnisse zu erzielen, sollen aus Sicht der Geoinformatik Mittel für die Objektidentifikation eingesetzt werden. Anstelle von Feldarbeit werden deshalb in der vorliegenden Arbeit multispektrale Fernerkundungsdaten als Primärdaten verwendet. Konkrete natürliche Objekte werden GIS-gestützt und automatisiert über große Flächen und Objektdichten aus Primärdaten identifiziert und charakterisiert. Im Rahmen der vorliegenden Arbeit wird eine automatisierte Prozesskette zur Objektidentifikation konzipiert. Es werden neue Ansätze und Konzepte der objektbasierten Identifikation von natürlichen isolierten terrestrischen Oberflächenformen entwickelt und implementiert. Die Prozesskette basiert auf einem Konzept, das auf einem generischen Ansatz für automatisierte Objektidentifikation aufgebaut ist. Die Prozesskette kann anhand charakteristischer quantitativer Parameter angepasst und so umgesetzt werden, womit das Konzept der Objektidentifikation modular und skalierbar wird. Die modulbasierte Architektur ermöglicht den Einsatz sowohl einzelner Module als auch ihrer Kombination und möglicher Erweiterungen. Die eingesetzte Methodik der Objektidentifikation und die daran anschließende Charakteristik der (geo)morphometrischen und morphologischen Parameter wird durch statistische Verfahren gestützt. Diese ermöglichen die Vergleichbarkeit von Objektparametern aus unterschiedlichen Stichproben. Mit Hilfe der Regressionsund Varianzanalyse werden Verhältnisse zwischen Objektparametern untersucht. Es werden funktionale Abhängigkeiten der Parameter analysiert, um die Objekte qualitativ zu beschreiben. Damit ist es möglich, automatisiert berechnete Maße und Indizes der Objekte als quantitative Daten und Informationen zu erfassen und unterschiedliche Stichproben anzuwenden. Im Rahmen dieser Arbeit bilden Thermokarstseen die Grundlage für die Entwicklungen und als Beispiel sowie Datengrundlage für den Aufbau des Algorithmus und die Analyse. Die Geovisualisierung der multivariaten natürlichen Objekte wird für die Entwicklung eines besseren Verständnisses der räumlichen Relationen der Objekte eingesetzt. Kern der Geovisualisierung ist das Verknüpfen von Visualisierungsmethoden mit kartenähnlichen Darstellungen. / The automated object identification represents a modern tool in geoinformatics (BLASCHKE et al., 2012). In order to achieve results in thematic mapping comparable among one another, considering geoinformatics, means of object identification should be applied. Therefore, instead of fieldwork, multispectral remote-sensing data have been used as a primary data source in this work. Specific natural objects have been GIS-based and automatically identified and characterised from the primary data over large areas and object densities. Within this work, an automated process chain for the object identification has been developed. New approaches and concepts of object-based identification of natural isolated terrestrial landforms have been developed and implemented. The process chain is based on a concept that develops a generic approach to the automated object identification. This process chain can be customised for and applied to specific objects by settings of characteristic quantitative parameters, by which the concept of object identification becomes modular and scalable. The modul-based architecture enables use of individual moduls as well as their combinations and possible expansions. The introduced methodology of object identification and the connected characteristics of (geo)morphometric and morphologic parameters has been supported by a static procedures. These enable the comparability of object parameters from different samples. With the help of regression and variance analysis, relations between object parameters have been explored. Functional dependencies of parameters have been analysed in order to qualitatively describe the objects. As a result, automatically computed dimensions and indices of the objects can be captured as quantitative data and informations an applied to varied samples. Within this work the thermokarst lakes represent the basis for the process development and an example and a data basis for the design of the algorithm and analysis. The goevisualisation of multivariant natural objects has been applied to develop better understanding of their spatial relations. The essence of the geovisualisation is to link the methods of visualisation to map-like presentation.

No abstract available

This thesis is devoted to the problems of defining and developing the basic building blocks of an automated surveillance system. As its initial step, a background-modeling algorithm is described for segmenting moving objects from the background, which is capable of adapting to dynamic scene conditions, as well as determining shadows of the moving objects. After obtaining binary silhouettes for targets, object association between consecutive frames is achieved by a hypothesis-based tracking method. Both of these tasks provide basic information for higher-level processing, such as activity analysis and object identification. In order to recognize the nature of an event occurring in a scene, hidden Markov models (HMM) are utilized. For this aim, object trajectories, which are obtained through a successful track, are written as a sequence of flow vectors that capture the details of instantaneous velocity and location information. HMMs are trained with sequences obtained from usual motion patterns and abnormality is detected by measuring the distance to these models. Finally, MPEG-7 visual descriptors are utilized in a regional manner for object identification. Color structure and homogeneous texture parameters of the independently moving objects are extracted and classifiers, such as Support Vector Machine (SVM) and Bayesian plug-in (Mahalanobis distance), are utilized to test the performance of the proposed person identification mechanism. The simulation results with all the above building blocks give promising results, indicating the possibility of constructing a fully automated surveillance system for the future.

Saab använder sig i nuläget av ett utdaterat system för att utföra tester av deras produkter. Systemet filmar ur olika vinklar och sammanfogar videoströmmarna till en slutgiltig video, där de sedan kan analysera resultatet av produkten. Enkortsdatorer är något som på senare år har blivit mer och mer populärt, Saab vill därför undersöka om det går att ersätta det äldre systemet med enkortsdatorer och kameror.Det ska undersökas om enkortsdatorn BeagleBoard klarar av att köra objektidentifiering samtidigt som den filmar och utför operationer som videosynkning, videokodning samt sparar den synkade filmen.Undersökningen visade att BeagleBoardens processor inte är tillräckligt kraftfull för att klara av objektidentifieringen utan hårdvarustöd. Istället behöver det utföras av en dator som bearbetar filmen i efterhand och plockar ut objekt. Det har förslagits en bättre metod för att göra objektidentifieringen smartare och lärande som kommer fungera bättre i Saabs fall. / Saab is currently using an old and complex system to perform tests of their products. The system is based on filming from different angles which will be merged to one film from which Saab can analyze the results of their products. Single-board computers is something that have become increasingly popular in the recent years, therefore, we are to investigate whether it is possible or not to replace the older systems with SBCs and cameras.We will also investigate whether the BeagleBoard is capable of detecting objects while filming, synchronizing, encoding and saving the video for later use.The result showed that the processor isn’t powerful enough to handle object identification without full hardware support. Instead, it needs to be performed afterwards by a computer which will identify objects in the video. A better method has been proposed to make object identification smarter and learning, which will work better in Saab’s case and their future work.

Mastitis is a common disease among cows in dairy farms. Diagnosis of the infection is today done manually, by analyzing bacteria growth on agar plates. However, classifiers are being developed for automated diagnostics using images of agar plates. Input images need to be of reasonable quality and consistent in terms of scale, positioning, perspective, and rotation for accurate classification. Therefore, this thesis investigates if a combination of image processing techniques can be used to match each input image to a pre-defined reference model. A method was proposed to identify important key points needed to register the input image to the reference model. The key points were defined by identifying the agar plate, its compartments, and its rotation within the image. The results showed that image registration with the correct key points was sufficient enough to match images of agar plates to a reference model despite any varieties in scale, position, perspective, or rotation. However, the accuracy depended on the identification of the salient features of the agar plate. Ultimately, the work proposes an approach using image registration to transform images of agar plates based on a pre-defined reference model, rather than a reference image.