本論文的主要目的是建立一套人臉辨識系統，即使在光源不足或甚至是完全黑暗的環境下也能夠正確地進行身分辨識。在完全黑暗的情形下，我們可以利用具有夜視功能(近紅外線)的攝影機來擷取影像，然而，近紅外線影像通常呈現亮度非常不均勻的情形，導致我們無法直接利用現有的人臉辨識系統來作辨識。因此，我們首先觀察近紅外線影像的特性，然後依據此特性提出一個影像成像的模型；接著，利用同構增晰的原理來減低因成像過程所造成的不均勻現象；經由實驗的結果，我們得知現有的全域式人臉辨識系統無法有效地處理近紅外線影像，因此，我們提出了一個新的區域式的人臉辨識演算法，針對光線不足的情況作特殊考量，以得到較佳的辨識結果。本論文實作的系統是以最近點分類法來作身份辨識，在現有的32個人臉影像資料集中，正確的辨識率達75%。 / The main objective of this thesis is to develop a face recognition system that could recognize human faces even when the surrounding environment is totally dark. The images of objects in total darkness can be captured using a relatively low-cost camcorder with the NightShot® function. By overcoming the illumination factor, a face recognition system would continue to function independent of the surrounding lighting condition. However, images acquired exhibit non-uniformity due to irregular illumination and current face recognition systems may not be put in use directly. In this thesis, we first investigate the characteristics of NIR images and propose an image formation model. A homomorphic processing technique built upon the image model is then developed to reduce the artifact of the captured images. After that, we conduct experiments to show that existing holistic face recognition systems perform poorly with NIR images. Finally, a more robust feature-based method is proposed to achieve better recognition rate under low illumination. A nearest neighbor classifier using Euclidean distance function is employed to recognize familiar faces from a database. The feature-based recognition method we developed achieves a recognition rate of 75% on a database of 32 people, with one sample image for each subject.

This thesis presents the design and evaluation of the location of a cell phone user, to enable more effective performance monitoring. One of the end-uses I propose is in emergency management, by means of a framework that distributes its functionality between establishing data-set characteristics that are relevant to the problem and a visual tool to evaluate resource-scheduling proposals. Manikarnika is a modular framework, which finds translation in a prototype for Reverse 111. The first steps in the process were to establish whether the parameters I hypothesized as useful, indeed were. Using a statistically significant amount of traces, obtained from real calls placed on the network, the utility of the location metric was established. In order to investigate utilizing a second metric of reputation, a benchmark for evaluating ideas from Social Networks research was proposed, in order to move from arbitrary testing to a more systematic environment. This dissertation details the measurement, design and evaluation of an end-to-end and modular framework for Emergency Management, where the functionality is distributed in order to easily incorporate the changing parameters of sources of information, emergency events, resource requirements of these events and identifying callers that might be able to provide better insight into a situation that is essentially very dynamic. The chasm between research proposals for various end-uses and the application of the same to real life is one that I have tried to bridge in my work. By incorporating pieces from core Electrical Engineering measurements and simulation and extending the use of what was originally a tool built for training Emergency Responders to analyze various resource scheduling agents, which take into account a diversity of administrative domains, I lay the ground work for one possible solution, Reverse 111, which proposes the use of proactive crowd-sourcing for emergency response, with easy extensions to commercial location-based applications.

e-Learning is an effective medium for delivering knowledge and skills. In spite of improvements in electronic delivery technologies, e-Learning is still a long way away from offering anything close to efficient and effective learning environments. To improve e-Learning experiences, much literature supports simulation based e-Learning. This thesis begins identifying various types of simulation models and their features that induce experiential learning. We focus on designing and constructing an easy-to-use Discrete Event Simulation (DES) tool for building engaging and informative interactive DES models that allow learners to control the models’ parameters and visualizations through runtime interactions. DES has long been used to support analysis and design of complex systems but its potential to enhance learning has not yet been fully utilized. We first present an application framework and its resulting classes for better structuring DES models. However, importing relevant classes, establishing relationships between their objects and representing lifecycles of various types of active objects in a language that does not support concurrency demand a significant cognitive workload. To improve this situation, we utilize two design patterns to ease model structuring and logic representation (both in time and space) through a drag and drop component approach. The patterns are the Delegation Event Model, used for linking between components and delegating tasks of executing and updating active objects’ lifecycles, and the MVC (Model-View-Controller) pattern, used for connecting the components to their graphical instrumentations and GUIs. Components implementing both design patterns support the process-oriented approach, can easily be tailored to store model states and visualizations, and can be extended to design higher level models through hierarchical simulation development. Evaluating this approach with both teachers and learners using ActionScript as an implementation language in the Flash environment shows that the resulting components not only help model designers with few programming skills to construct DES models, but they also allow learners to conduct various experiments through interactive GUIs and observe the impact of changes to model behaviour through a range of engaging visualizations. Such interactions can motivate learners and make their learning an enjoyable experience.

No description available.

No description available.

No description available.

No description available.

No description available.

Radical polymerisation (RP) has for many years been an industrially important process, and the kinetics of the process remains an active area of research. As polymerisation proceeds, converting monomer (small molecules) into polymer (long chain molecules), chemical species of a variety of chain lengths are produced. Recent work has pointed toward the fact that rate coefficients for polymerisation reactions (specifically, termination and propagation) are often dependent on the chain-length of the reacting species. The focus of this thesis is to study the effects of chain-length-dependent reactions on the kinetics of RP, by using computer-based modeling and comparing the results of such modeling with experimental data. This enables the understanding of otherwise inexplicable trends and the building of more mechanistically detailed and accurate models for RP kinetics. In Chapter 2, a new model for termination is developed, connecting observations and analyses of termination kinetics at short chain lengths (particularly small molecule studies) with other observations and analyses at long chain lengths (conventional RP kinetics studies) in order to construct a model for termination that is shown to be capable of coherently describing termination kinetics at any chain length. In Chapter 3, this new model for termination is tested at short chain lengths on polymerisations with large quantities of added chain transfer agent. With the inclusion of chain-length-dependent propagation in the model, the model for termination is validated. Chapter 4 is aimed at extending an existing reduced-variable, compact, 'universal' description of steady-state RP kinetics by incorporating all known chain-length dependent reactivities. This both increases computational efficiency over other approaches and provides easily evaluated, approximate analytical expressions for RP kinetics. This foundational theory is applied in Chapter 5 to reach a deeper understanding of the behaviour of the model, and show how experimental data may readily be analysed to extract information about chain-length-dependent termination kinetics. In Chapter 6, the effect of chain-length dependent reactivities on the important technique of single-pulse pulsed-laser polymerisation is investigated, and this technique is validated as the best experimental method for investigation of termination kinetics. In general, a central result of this thesis is that chain-length-dependent reactivities, when acknowledged and properly incorporated into models, can explain many phenomena in RP kinetics which otherwise seem difficult to account for. No exceptions to this principle have been found.

Section 1: Introduction, Section 2: Physical Features, Section 3: Biological Resources, Section 4: Social/Economic Characteristics, Section 5: Important Resources, Section 6: Watershed Classification, Section 7: Watershed Management, Section 8: Local Watershed Planning, Section 9: Key Elements, Appendix A: Table 1, Appendix B: Suggested Readings, Appendix C: RUSLE, Appendix D: AGWA