Mathematical morphology was first applied to binary images and readily extended to grey-level images. In extending mathematical morphology to colour it is difficult to define a suitable unambiguous ordering. We present two complete ordering schemes based on colour difference and similarity ordering for colour morphology. A novel colour difference formula is first introduced. This colour difference formula is based on colour extrema derived from a simple physical model of image formation and avoids the more arbitrary mathematical and perceptual definitions previously reported. Moreover, we define similarity criteria as the basis for mathematical morphology that can be used with flat and non-flat structuring elements. The proposed orderings meet the properties of mathematical morphology, and provide a harmonised approach for binary, grey-level and colour morphology. A comparison of ordering schemes for dilation, erosion, opening, closing and filtering operator shows the colour difference-based ordering presented here to be at least as good as other ordering schemes and better than some of the well principled, previously reported methods in not generating artefacts and reducing image noise. Additionally, the development of a similarity-based ordering to perform morphological gradient and Hit-or-Miss transforms for colour images is presented.

The topic of this thesis is users' models: the representations users may form of the computer system which they are interacting with. It has been proposed that user interfaces which support the construction of appropriate users' models facilitate learning and use of computer systems. Users' models have been a topic of research in human-computer interaction (HCI) since 1984, but to date, no knowledge exists which could be applied by designers of computer systems. The aim of the thesis is to address this problem and contribute to the development of an integrated and applicable body of knowledge on users' models. The thesis commences with an examination of the history and current state of the discipline of human-computer interaction to establish the context and determine the appropriate methods for conducting research on users' models. Since mental representations and mental models originate from the related disciplines of psychology and cognitive science, the review of the literature starts with an outline of the relevant theories, followed by a review of the theoretical and empirical work to date on users' model in HCI. The review concludes that more exploratory empirical work is required to obtain data from which evidence for, and descriptions of, users' models could be derived; however, suitable methods for eliciting and describing users' models have to be devised first. The second part of the thesis describes a series of five observational studies of users interacting with application software. The studies employed different scenarios, ranging from traditional experimental-style scenarios, with users working through a series of tasks, to constructive interaction scenarios, where users interacted with a co-investigator playing the role of a learner or co-learner. All studies were recorded on video, transcribed and analysed. Advantages and drawbacks of the scenarios for eliciting users' models are identified and discussed. The analysis of the tapes and transcripts provides some evidence of users' models; the conclusions of the thesis provide an outline of how theories regarding users' models can be formulated, and tested, using the data collected.

This thesis describes work on defining and modelling text comprehension. The basis of my approach is a theory of comprehension as a form of abductive reasoning. The specific problem addressed is inferential control and the management of alternative, competing representations. This problem is related to issues of representation quality, with decisions between representations being made on the basis of quality comparisons. Simultaneously, monitoring of representation quality determines when comprehension halts; in other words, there is some kind of threshold against which quality is compared. In the first part of the thesis I analyse concepts of representation quality, describing the structure of episodic and semantic representations and processes. I then look at metrics for representation quality before developing my own metric. The metric is based on the concept of incoherence, derived from the structural potential of representations. The second part of the thesis describes a computational model of incoherence, the Incoherence-Driven Comprehender (IDC). IDC combines AI implementation technology with insights from cognitive psychological studies of text comprehension. I show how IDC can be applied to various comprehension tasks. Throughout the thesis I suggest how aspects of IDC's architecture and behaviour may offer a fresh perspective on human comprehension.

The enormous potential of cloud computing for improved and cost-effective service has generated unprecedented interest in its adoption. However, a potential cloud user faces numerous risks regarding service requirements, cost implications of failure and uncertainty about cloud providers’ ability to meet service level agreements. These risks hinder the adoption of cloud computing. We motivate the need for a new requirements engineering methodology for systematically helping businesses and users to adopt cloud services and for mitigating risks in such transition. The methodology is grounded in goal-oriented approaches for requirements engineering. We argue that Goal-Oriented Requirements Engineering (GORE) is a promising paradigm to adopt for goals that are generic and flexible statements of users’ requirements, which could be refined, elaborated, negotiated, mitigated for risks and analysed for economics considerations. The methodology can be used by small to large scale organisations to inform crucial decisions related to cloud adoption. We propose a risk management framework based on the principle of GORE. In this approach, we liken risks to obstacles encountered while realising cloud user goals, therefore proposing cloud-specific obstacle resolution tactics for mitigating identified risks. The proposed framework shows benefits by providing a principled engineering approach to cloud adoption and empowering stakeholders with tactics for resolving risks when adopting the cloud. We extend the work on GORE and obstacles for informing the adoption process. We argue that obstacles’ prioritisation and their resolution is core to mitigating risks in the adoption process. We propose a novel systematic method for prioritising obstacles and their resolution tactics using Analytical Hierarchy Process (AHP). To assess the AHP choice of the resolution tactics we support the method by stability and sensitivity analysis.

Following the success in 3D face recognition, the face processing community is now trying to establish good 3D facial expression recognition. Facial expressions provide the cues of communication in which we can interpret the mood, meaning and emotions at the same time. With current advanced 3D scanners technology, direct anthropometric measurements (i.e. the comparative study of sizes and proportions of the human body) are easily obtainable and it offers 3D geometrical data suitable for 3D face processing studies. Instead of using the raw 3D facial points, we extracted its derivative which gives us 3D facial surface normals. We constructed a statistical model for variations in facial shape due to changes in six basic expressions using 3D facial surface normals as the feature vectors. In particular, we are interested in how such facial expression variations manifest themselves in terms of changes in the field of 3D facial surface normals. We employed a modular approach where a module contains the facial features of a distinct facial region. The decomposition of a face into several modules promotes the learning of a facial local structure and therefore the most discriminative variation of the facial features in each module is emphasized. We decomposed a face into six modules and the expression classification for each module is carried out independently. We constructed a Weighted Voting Scheme (WVS) to infer the emotion underlying a collection of modules using a weight that is determined using the AdaBoost learning algorithm. Using our approach, using 3D facial surface normal as the feature vector of WVS yields a better performance than 3D facial points and 3D distance measurements in facial expression classification using both WVS and Majority Voting Scheme (MVS). The attained results suggest surface normals do indeed produce a comparable result particularly for six basic facial expressions with no intensity information.

The landscape of security has been changed by the increase in online market places, and the rapid growth of mobile and wireless networks. Users are now exposed to greater risks as they interact anonymously in these domains. Despite the existing security paradigms, trust among users remains a problem. Reputation systems have now gained popularity because of their effectiveness in providing trusted interactions. We argue that managing reputation by relying on history alone and/or biased opinions is inadequate for security, because such an approach exposes the domain to vulnerabilities. Alternatively, the use of historical, recent and anticipated events supports effective reputation management. We investigate how the dynamic data-driven application systems paradigm can aid reputation management. We suggest the use of the paradigm's primitives, which includes the use of controller and simulation components for performing computations and predictions. We demonstrate how a dynamic framework can provide effective reputation management that is not influenced by biased observations. This is an online decision support system that can enable stakeholders make informed judgments. To highlight the framework's usefulness, we report on its predictive performance through an evaluation stage. Our results indicate that a dynamic data-driven approach can lead to effective reputation management in trust-reliant domains.

Most approaches to spatial image management involve GPS or image processing. In this thesis, a sensor-focused alternative is explored. It requires user and camera tracking, particularly challenging in indoor environments. Possible indoor tracking methods are evaluated and pedestrian dead reckoning is selected. A study is conducted to evaluate sensors and choose a combination for pedestrian and camera tracking. Gyroscope and accelerometer offer comparable step detection performance, with gyroscope and tilt compensated compass providing heading data. Images taken from the same viewpoint are successfully arranged using panorama stitching without any image processing. The results compare favourably to conventional methods. While lacking visual definition of image processing methods, they can complement them if used in tandem. Sensor compositing and pedestrian tracking are implemented in a unified system. Several methods for fusing compass and gyroscope data are compared, but do not produce statistically significant improvement over using just the compass. The system achieves loop closure accuracy of 91% of path length and performs consistently across multiple participants. The final system can be used in GPS-denied locations and presents an image content independent way of managing photographs. It contributes to pedestrian tracking and image composting fields and has potential commercial uses (illustrated by an example Android app).

Metaheuristics are approximation optimisation techniques widely applied to solve complex optimisation problems. Despite a large number of developed metaheuristic algorithms, a limited amount of work has been done to understand on which kinds of problems the proposed algorithm will perform well or poorly and why. A useful solution to this dilemma is to use fitness landscape analysis to gain an in-depth understanding of which algorithms, or algorithm variants are best suited for solving which kinds of problem instances, even to dynamically determine the best algorithm configuration during different stages of a search algorithm. This thesis for the first time bridges the gap between fitness landscape analysis and algorithm configuration, i.e., finding the best suited configuration of a given algorithm for solving a particular problem instance. Studies in this thesis contribute to the following: a. Developing a novel and effective approach to characterise fitness landscapes and measure problem difficulty with respect to algorithms. b. Incorporating fitness landscape analysis in building a generic (problem-independent) approach, which can perform automatic algorithm configuration on a per-instance base, and in designing novel and effective algorithm configurations. c. Incorporating fitness landscape analysis in establishing a generic framework for designing adaptive heuristic algorithms.

A category CoLog of distributive laws is introduced to unify different approaches to modal logic for coalgebras, based merely on the presence of a contravariant functor P that maps a state space to its collection of predicates. We show that categorical constructions, including colimits, limits, and compositions of distributive laws as a tensor product, in CoLog generalise and extend existing constructions given for Set coalgebraic logics and that the framework does not depend on any particular propositional logic or state space. In the case that P establishes a dual adjunction with its dual functor S, we show that a canonically defined coalgebraic logic exists for any type of coalgebras. We further restrict our discussion to finitary algebraic logics and study equational coalgebraic logics. Objects of predicate liftings are used to characterise equational coalgebraic logics. The expressiveness problem is studied via the mate correspondence, which gives an isomorphism between CoLog and the comma category from the pre-composition to the post-composition with S. Then, the modularity of the expressiveness is studied in the comma category via the notion of factorisation system.

Every day, robots are being deployed in more challenging environments, where they are required to perform complex tasks. In order to achieve these tasks, robots rely on intelligent deliberation algorithms. In this thesis, we study two deliberation approaches – task scheduling and task planning. We extend these approaches in order to not only deal with temporal and spatial constraints imposed by the environment, but also exploit them to be more efficient than the state-of-the-art approaches. Our first main contribution is a scheduler that exploits a heuristic based on Allen’s interval algebra to prune the search space to be traversed by a mixed integer program. We empirically show that the proposed scheduler outperforms the state of the art by at least one order of magnitude. Furthermore, the scheduler has been deployed on several mobile robots in long-term autonomy scenarios. Our second main contribution is the POPMERX algorithm, which is based on merging of partially ordered temporal plans. POPMERX first reasons with the spatial and temporal structure of separately generated plans. Then, it merges these plans into a single final plan, while optimising the makespan of the merged plan. We empirically show that POPMERX produces better plans that the-state-ofthe- art planners on temporal domains with time windows.