Notas / Resumen

For continuous-time descriptor systems with all system matrices incorporated with norm-bounded uncertainties, this thesis addresses robust admissibility and robust H¡Û analysis and the related state feedback design. The results are further extended to systems with time-varying state delay within a known interval. The former part of the thesis extends the current research of considering uncertainty only at the state derivative matrix to the case uncertainty being assumed at all system matrices. While the latter part of the thesis extends the current research in two folds: the state derivative matrix is allowed to be uncertain and the delay is allowed to be time-varying. Since all the results are derived in the LMI-based framework, examples with efficient numerical verifications are included to illustrate the derived results.

For linear time-invariant descriptor models, this dissertation studies the extended strictly positive real (ESPR) design of continuous-time systems and the strictly positive real (SPR) analysis and design of discrete-time systems, respectively, all in the LMI framework. For a continuous-time system, by the LMI-based ESPR Lemma, a controller is designed such that the closed-loop system has its transfer matrix being ESPR while admissibility of the compensated descriptor system is guaranteed. Three forms of synthesis are considered, i.e. the static state feedback synthesis, estimated state feedback synthesis, and the dynamic output feedback synthesis. Moreover, design criterion of a dynamic output feedback controller in the state-space model is also addressed. For a discrete-time system, an LMI-based SPR characterization is developed. After giving the definition of SPR, the Cayley transformation is used to establish formulas bridging the admissible realizations for SPR and strictly bounded real (SBR) transfer matrices. Based on them, an LMI-based necessary and sufficient condition for a descriptor system to be, simultaneously, admissible and SPR is derived. When the descriptor variables are transformed into the SVD coordinate, it is shown that such a condition will have solution in the block diagonal form. Based on this result, the problem of static state feedback design to make transfer matrix of the closed-loop systems SPR is tackled. The problems of robust ESPR and SPR analysis and design when the considered systems have norm-bounded unstructured uncertainty are also addressed. Similarly, LMI-based conditions to guarantee robust admissibility with transfer matrices being ESPR for continuous systems or being SPR for discrete systems are proposed. Based on them, for continuous systems, a static state feedback controller and a dynamic output feedback controller are designed to make the entire family of uncertain closed-loop systems robustly admissible with transfer matrices being ESPR. While for discrete systems, only static state feedback controller is designed to achieve the robust admissibility and robust SPR property. Finally, based on ESPR lemma (or SPR lemma), we propose a new LMI-based robust admissibility analysis for a class of LTI continuous-time (or discrete-time) descriptor systems with convex polytopic uncertainties appearing on all the system matrices. Moreover, the development of state feedback controllers stemmed from these analysis results is also investigated. It is shown that the provided method has the capability to tackle the problem of computing a required feedback gain matrix for systems with either constant or polytopically dependent derivative (or advanced) state matrix in a unified way. Besides, the application of SPR property to absolute stability problem involving an LTI discrete-time descriptor system and a memoryless time-varying nonlinearity is also addressed. Since all conditions are expressed in LMIs, the obtained results are numerically tractable. It is illustrated by several numerical examples.

Local descriptors are increasingly used for the task of object recognition because of their perceived robustness with respect to occlusions and to global geometrical deformations. Such a descriptor--based on a set of oriented Gaussian derivative filters-- is used in our recognition system. We report here an evaluation of several techniques for orientation estimation to achieve rotation invariance of the descriptor. We also describe feature selection based on a single training image. Virtual images are generated by rotating and rescaling the image and robust features are selected. The results confirm robust performance in cluttered scenes, in the presence of partial occlusions, and when the object is embedded in different backgrounds.

Based on the Lyapunov stability theorem, two different variable structure controllers are proposed in this thesis for two different classes of multi-variable descriptor systems subject to matched nonlinear perturbations. The integral variable structure controller is proposed first for solving the stabilization problems, and model reference variable structure controller is the second for solving the state tracking problems. Both proposed control schemes can guarantee the trajectories of the controlled systems to lie in the sliding surface from initial time, so that the properties of regularity, impulse free, and stability can be obtained. Two numerical examples are given for demonstrating the feasibility of the proposed control schemes.

In a time when cattle herds grow continually larger the need for automatic methods to detect diseases is ever increasing. One possible method to discover diseases is to use thermal images and automatic head and eye detectors. In this thesis an eye detector and a head detector is implemented using the Random Forests classifier. During the implementation the classifier is evaluated using three different descriptors: Histogram of Oriented Gradients, Local Binary Patterns, and a descriptor based on pixel differences. An alternative classifier, the Support Vector Machine, is also evaluated for comparison against Random Forests. The thesis results show that Histogram of Oriented Gradients performs well as a description of cattle heads, while Local Binary Patterns performs well as a description of cattle eyes. The provided descriptor performs almost equally well in both cases. The results also show that Random Forests performs approximately as good as the Support Vector Machine, when the Support Vector Machine is paired with Local Binary Patterns for both heads and eyes. Finally the thesis results indicate that it is easier to detect and locate cattle heads than it is to detect and locate cattle eyes. For eyes, combining a head detector and an eye detector is shown to give a better result than only using an eye detector. In this combination heads are first detected in images, followed by using the eye detector in areas classified as heads.

Local descriptors are increasingly used for the task of object recognition because of their perceived robustness with respect to occlusions and to global geometrical deformations. Such a descriptor--based on a set of oriented Gaussian derivative filters-- is used in our recognition system. We report here an evaluation of several techniques for orientation estimation to achieve rotation invariance of the descriptor. We also describe feature selection based on a single training image. Virtual images are generated by rotating and rescaling the image and robust features are selected. The results confirm robust performance in cluttered scenes, in the presence of partial occlusions, and when the object is embedded in different backgrounds.

OBJECTIVES: During health guideline development, panel members often have implicit, different definitions of health outcomes that can lead to variability in evidence synthesis and recommendations. McMaster GRADE Centre researchers developed a standardized description of health outcomes using the health marker state format. We aimed to determine which aspects of the development, content, and use of marker states were valuable to guideline developers. STUDY DESIGN & SETTING: We conducted a case study of marker state development with the European Commission Initiative on Breast Cancer (ECIBC) Guidelines Development Group (GDG). Eighteen GDG members provided written and interview feedback on the process. Using the health marker states, 2 health utility rating surveys were conducted near the beginning and end of development respectively. RESULTS: We developed 24 marker states for outcomes related to breast cancer screening and diagnosis. Feedback from GDG members revealed that marker states could be useful for developing recommendations and improving transparency of guideline methods. Comparison of the two health utility surveys showed a decrease in standard deviation in the second survey across 21 (88%) of the outcomes. CONCLUSIONS: Health marker states are a promising method, satisfying the pre-requisite of being feasible, acceptable, and with some initial result on reduction of variance of health utility scores. / Thesis / Master of Public Health (MPH) / OBJECTIVES: During health guideline development, panel members often have implicit, different definitions of health outcomes that can lead to variability in evidence synthesis and recommendations. McMaster GRADE Centre researchers developed a standardized description of health outcomes using the health marker state format. We aimed to determine which aspects of the development, content, and use of marker states were valuable to guideline developers. STUDY DESIGN & SETTING: We conducted a case study of marker state development with the European Commission Initiative on Breast Cancer (ECIBC) Guidelines Development Group (GDG). Eighteen GDG members provided written and interview feedback on the process. Using the health marker states, 2 health utility rating surveys were conducted near the beginning and end of development respectively. RESULTS: We developed 24 marker states for outcomes related to breast cancer screening and diagnosis. Feedback from GDG members revealed that marker states could be useful for developing recommendations and improving transparency of guideline methods. Comparison of the two health utility surveys showed a decrease in standard deviation in the second survey across 21 (88%) of the outcomes. CONCLUSIONS: Health marker states are a promising method, satisfying the pre-requisite of being feasible, acceptable, and with some initial result on reduction of variance of health utility scores.

Compiler-Inserted Runtime Adaptation (CIRA) is a compilation and runtime adaptation strategy which has great potential for increasing performance in multicore systems. In this strategy, the compiler inserts directives into the application which will adapt the application at runtime. Its ability to overcome the obstacles of architectural and environmental diversity coupled with its flexibility to work with many programming languages and styles of applications make it a very powerful tool. However, it is not complete. In fact, there are many pieces still needed to accomplish these lofty goals. This work describes the automatic discovery of parallelism inherent in an application and the generation of an intermediate representation to expose that parallelism. This work shows on six benchmark applications that a significant amount of parallelism which was not initially apparent can be automatically discovered. This work also shows that the parallelism can then be exposed in a representation which is also automatically generated. This is accomplished by a series of analysis and transformation passes with only minimal programmer-inserted directives. This series of passes forms a necessary part of the CIRA toolchain called the concurrency compiler. This concurrency compiler proves that a representation with exposed parallelism and locality can be generated by a compiler. It also lays the groundwork for future, more powerful concurrency compilers. This work also describes the extension of the intermediate representation to support hierarchy, a prerequisite characteristic to the creation of the concurrency compiler. This extension makes it capable of representing many more applications in a much more effective way. This extension to support hierarchy allows much more of the parallelism discovered by the concurrency compiler to be stored in the representation.

Every day we recognize a numerous objects and human brain can recognize objects under many conditions. The way in which humans are able to identify an object is remarkably fast even in different size, colours or other factors. Computers or robots need computational tools to identify objects. Shape descriptors are one of the tools commonly used in image processing applications. Shape descriptors are regarded as mathematical functions employed for investigating image shape information. Various shape descriptors have been studied in the literature. The aim of this thesis is to develop new shape descriptors which provides a reasonable alternative to the existing methods or modified to improve them. Generally speaking shape descriptors can be categorized into various taxonomies based on the information they use to compute their measures. However, some descriptors may use a combination of boundary and interior points to compute their measures. A new shape descriptor, which uses both region and contour information, called centeredness measure has been defined. A new alternative ellipticity measure and sensitive family ellipticity measures are introduced. Lastly familiy of ellipticity measures, which can distinguish between ellipses whose ratio between the length of the major and minor axis differs, have been presented. These measures can be combined and applied in different image processing applications such as image retrieval and classification. This simple basis is demonstrated through several examples.