Doctoral Degrees (Computer Science)

Permanent URI for this collectionhttps://hdl.handle.net/10413/7113

Browse

Now showing 1 - 20 of 38

The adoption of Web 2.0 tools in teaching and learning by in-service secondary school teachers: the Mauritian context.
(2018) Pyneandee, Marday.; Govender, Desmond Wesley.; Oogarah-Pratap, Brinda.
With the current rapid increase in use of Web 2.0 tools by students, it is becoming necessary for teachers to understand what is happening in this social networking phenomenon, so that they can better understand the new spaces that students inhabit and the implications for students’ learning and investigate the wealth of available Web 2.0 tools, and work to incorporate some into their pedagogical and learning practices. Teachers are using the Internet and social networking tools in their personal lives. However, there is little empirical evidence on teachers’ viewpoints and usage of social media and other online technologies to support their classroom practice. This study stemmed from the urgent need to address this gap by exploring teachers’ perceptions, and experience of the integration of online technologies, social media, in their personal lives and for professional practice to find the best predictors of the possibility of teachers’ using Web 2.0 tools in their professional practice. Underpinning the study is a conceptual framework consisting of core ideas found in the unified theory of acceptance and use of technology (UTAUT) and technology pedagogy and content knowledge (TPACK) models. The conceptual framework, together with a review of relevant literature, enabled the formulation of a theoretical model for understanding teachers’ intention to exploit the potential of Web 2.0 tools. The model was then further developed using a mixed-method, two-phase methodology. In the first phase, a survey instrument was designed and distributed to in-service teachers following a Postgraduate Certificate in Education course at the institution where the researcher works. Using the data collected from the survey, exploratory factor analysis, correlational analysis and multiple regression analysis were used to refine the theoretical model. Other statistical methods were also used to gain further insights into teachers’ perceptions of use of Web 2.0 tools in their practices. In the second phase of the study, survey respondents were purposefully selected, based on quantitative results, to participate in interviews. The qualitative data yielded from the interviews was used to support and enrich understanding of the quantitative findings. The constructs teacher knowledge and technology pedagogy knowledge from the TPACK model and the constructs effort expectancy, facilitating conditions and performance expectancy are the best predictors of teachers’ intentions to use Web 2.0 tools in their professional practice. There was an interesting finding on the relationship between UTAUT and TPACK constructs. The constructs performance expectancy and effort expectancy had a significant relationship with all the TPACK constructs – technology knowledge, technology pedagogy knowledge, pedagogical content knowledge (PCK), technology and content knowledge and TPACK – except for content knowledge and pedagogical knowledge. The association between the TPACK construct PCK with the UTAUT constructs performance expectancy and effort expectancy was an unexpected finding because PCK is only about PCK and has no technology component. The theoretical contribution of this study is the model, which is teachers’ intention of future use of Web 2.0 tools in their professional practice. The predictive model, together with other findings, enhances understanding of the nature of teachers’ intention to utilise Web 2.0 tools in their professional practice. Findings from this study have implications for school infrastructure, professional development of teachers and an ICT learning environment to support the adoption of Web 2.0 tools in teaching practices and are presented as guiding principles at the end of the study.
Automated design of genetic programming of classification algorithms.
(2018) Nyathi, Thambo.; Pillay, Nelishia.
Over the past decades, there has been an increase in the use of evolutionary algorithms (EAs) for data mining and knowledge discovery in a wide range of application domains. Data classification, a real-world application problem is one of the areas EAs have been widely applied. Data classification has been extensively researched resulting in the development of a number of EA based classification algorithms. Genetic programming (GP) in particular has been shown to be one of the most effective EAs at inducing classifiers. It is widely accepted that the effectiveness of a parameterised algorithm like GP depends on its configuration. Currently, the design of GP classification algorithms is predominantly performed manually. Manual design follows an iterative trial and error approach which has been shown to be a menial, non-trivial time-consuming task that has a number of vulnerabilities. The research presented in this thesis is part of a large-scale initiative by the machine learning community to automate the design of machine learning techniques. The study investigates the hypothesis that automating the design of GP classification algorithms for data classification can still lead to the induction of effective classifiers. This research proposes using two evolutionary algorithms,namely,ageneticalgorithm(GA)andgrammaticalevolution(GE)toautomatethe design of GP classification algorithms. The proof-by-demonstration research methodology is used in the study to achieve the set out objectives. To that end two systems namely, a genetic algorithm system and a grammatical evolution system were implemented for automating the design of GP classification algorithms. The classification performance of the automated designed GP classifiers, i.e., GA designed GP classifiers and GE designed GP classifiers were compared to manually designed GP classifiers on real-world binary class and multiclass classification problems. The evaluation was performed on multiple domain problems obtained from the UCI machine learning repository and on two specific domains, cybersecurity and financial forecasting. The automated designed classifiers were found to outperform the manually designed GP classifiers on all the problems considered in this study. GP classifiers evolved by GE were found to be suitable for classifying binary classification problems while those evolved by a GA were found to be suitable for multiclass classification problems. Furthermore, the automated design time was found to be less than manual design time. Fitness landscape analysis of the design spaces searched by a GA and GE were carried out on all the class of problems considered in this study. Grammatical evolution found the search to be smoother on binary classification problems while the GA found multiclass problems to be less rugged than binary class problems.
Automatic dental caries detection in bitewing radiographs.
(2022) Majanga, Vincent Idah.; Viriri, Serestina.
Dental Caries is one of the most prevalent chronic disease around the globe. Distinguishing carious lesions has been a challenging task. Conventional computer aided diagnosis and detection methods in the past have heavily relied on visual inspection of teeth. These are only effective on large and clearly visible caries on affected teeth. Conventional methods have been limited in performance due to the complex visual characteristics of dental caries images, which consists of hidden or inaccessible lesions. Early detection of dental caries is an important determinant for treatment and benefits much from the introduction of new tools such as dental radiography. A method for the segmentation of teeth in bitewing X-rays is presented in this thesis, as well as a method for the detection of dental caries on X-ray images using a supervised model. The diagnostic method proposed uses an assessment protocol that is evaluated according to a set of identifiers obtained from a learning model. The proposed technique automatically detects hidden and inaccessible dental caries lesions in bitewing radio graphs. The approach employed data augmentation to increase the number of images in the data set in order to have a total of 11,114 dental images. Image pre-processing on the data set was through the use of Gaussian blur filters. Image segmentation was handled through thresholding, erosion and dilation morphology, while image boundary detection was achieved through active contours method. Furthermore, the deep learning based network through the sequential model in Keras extracts features from the images through blob detection. Finally, a convexity threshold value of 0.9 is introduced to aid in the classification of caries as either present or not present. The relative efficacy of the supervised model in diagnosing dental caries when compared to current systems is indicated by the results detailed in this thesis. The proposed model achieved a 97% correct diagnostic which proved quite competitive with existing models.
Bi-modal biometrics authentication on iris and signature.
(2010) Viriri, Serestina.; Tapamo, Jules-Raymond.
Multi-modal biometrics is one of the most promising avenues to address the performance problems in biometrics-based personal authentication systems. While uni-modal biometric systems have bolstered personal authentication better than traditional security methods, the main challenges remain the restricted degrees of freedom, non-universality and spoof attacks of the traits. In this research work, we investigate the performance improvement in bi-modal biometrics authentication systems based on the physiological trait, the iris, and behavioral trait, the signature. We investigate a model to detect the largest non-occluded rectangular part of the iris as a region of interest (ROI) from which iris features are extracted by a cumulative-sums-based grey change analysis algorithm and Gabor Filters. In order to address the space complexity of biometric systems, we proposed two majority vote-based algorithms which compute prototype iris features codes as the reliable specimen templates. Experiments obtained a success rate of 99.6%. A text-based directional signature verification algorithm is investigated. The algorithm verifies signatures, even when they are composed of symbols and special unconstrained cursive characters which are superimposed and embellished. The experimental results show that the proposed approach has an improved true positive rate of 94.95%. A user-specific weighting technique, the user-score-based, which is based on the different degrees of importance for the iris and signature traits of an individual, is proposed. Then, an intelligent dual ν-support vector machine (2ν-SVM) based fusion algorithm is used to integrate the weighted match scores of the iris and signature modalities at the matching score level. The bi-modal biometrics system obtained a false rejection rate (FRR) of 0.008, and a false acceptance rate (FAR) of 0.001.
Deep learning for brain tumor segmentation and survival prediction.
(2024) Magadza, Tirivangani Batanai Hendrix Takura.; Viriri, Serestina.
A brain tumor is an abnormal growth of cells in the brain that multiplies uncontrolled. The death of people due to brain tumors has increased over the past few decades. Early diagnosis of brain tumors is essential in improving treatment possibilities and increasing the survival rate of patients. The life expectancy of patients with glioblastoma multiforme (GBM), the most malignant glioma, using the current standard of care is, on average, 14 months after diagnosis despite aggressive surgery, radiation, and chemotherapies. Despite considerable efforts in brain tumor segmentation research, patient diagnosis remains poor. Accurate segmentation of pathological regions may significantly impact treatment decisions, planning, and outcome monitoring. However, the large spatial and structural variability among brain tumors makes automatic segmentation a challenging problem, leaving brain tumor segmentation an open challenge that warrants further research endeavors. While several methods automatically segment brain tumors, deep learning methods are becoming widespread in medical imaging due to their resounding performance. However, the boost in performance comes at the cost of high computational complexity. Therefore, to improve the adoption rate of computer-assisted diagnosis in clinical setups, especially in developing countries, there is a need for more computational and memoryefficient models. In this research, using a few computational resources, we explore various techniques to develop deep learning models accurately for segmenting the different glioma sub-regions, namely the enhancing tumor, the tumor core, and the whole tumor. We quantitatively evaluate the performance of our proposed models against the state-of-the-art methods using magnetic resolution imaging (MRI) datasets provided by the Brain Tumor Segmentation (BraTS) Challenge. Lastly, we use segmentation labels produced by the segmentation task and MRI multimodal data to extract appropriate imaging/radiomic features to train a deep learning model for overall patient survival prediction.
Deep learning framework for speech emotion classification.
(2024) Akinpelu, Samson Adebisi.; Viriri, Serestina.
A robust deep learning-based approach for the recognition and classification of speech emotion is proposed in this research work. Emotion recognition and classification occupy a conspicuous position in human-computer interaction (HCI) and by extension, determine the reasons and justification for human action. Emotion plays a critical role in decision-making as well. Distinguishing among various emotions (angry, sad, happy, neutral, disgust, fear, and surprise) that exist from speech signals has however been a long-term challenge. There have been some limitations associated with existing deep learning techniques as a result of the complexity of features from human speech (sequential data) which consists of insufficient label datasets, Noise and Environmental Factors, Cross-cultural and Linguistic Differences, Speakers’ Variability and Temporal Dynamics. There is also a heavy reliance on huge parameter tunning, especially for millions of parameters before the model can learn the expected emotional features necessary for classification emotion, which often results in computational complexity, over-fitting, and poor generalization. This thesis presents an innovative deep learning framework-based approach for the recognition and classification of speech emotions. The deep learning techniques currently in use for speech-emotion classification are exhaustively and analytically reviewed in this thesis. This research models various approaches and architectures based on deep learning to build a framework that is dependable and efficient for classifying emotions from speech signals. This research proposes a deep transfer learning model that addresses the shortcomings of inadequate training datasets for the classification of speech emotions. The research also models advanced deep transfer learning in conjunction with a feature selection algorithm to obtain more accurate results regarding the classification of speech emotion. Speech emotion classification is further enhanced by combining the regularized feature selection (RFS) techniques and attention-based networks for the classification of speech emotion with a significant improvement in the emotion recognition results. The problem of misclassification of emotion is alleviated through the selection of salient features that are relevant to emotion classification from speech signals. By combining regularized feature selection with attention-based mechanisms, the model can better understand emotional complexities and outperform conventional ML model emotion detection algorithms. The proposed approach is very resilient to background noise and cultural differences, which makes it suitable for real-world applications. Having investigated the reasons behind the enormous computing resources required for many deep learning based methods, the research proposed a lightweight deep learning approach that can be deployed on low-memory devices for speech emotion classification. A redesigned VGGNet with an overall model size of 7.94MB is utilized, combined with the best-performing classifier (Random Forest). Extensive experiments and comparisons with other deep learning models (DenseNet, MobileNet, InceptionNet, and ResNet) over three publicly available speech emotion datasets show that the proposed lightweight model improves the performance of emotion classification with minimal parameter size. The research further devises a new method that minimizes computational complexity using a vision transformer (ViT) network for speech emotion classification. The ViT model’s capabilities allow the mel-spectrogram input to be fed into the model, allowing for the capturing of spatial dependencies and high-level features from speech signals that are suitable indicators of emotional states. Finally, the research proposes a novel transformer model that is based on shift-window for efficient classification of speech emotion on bi-lingual datasets. Because this method promotes feature reuse, it needs fewer parameters and works well with smaller datasets. The proposed model was evaluated using over 3000 speech emotion samples from the publicly available TESS, EMODB, EMOVO, and bilingual TESS-EMOVO datasets. The results showed 98.0%, 98.7%, and 97.0% accuracy, F1-Score, and precision, respectively, across the 7 classes of emotion.
The design and simulation of routing protocols for mobile ad hoc networks.
(2000) Kabeto, Mieso Denko.; Goddard, Wayne.
This thesis addresses a novel type of network known as a mobile ad hoc network. A mobile ad hoc network is a collection of entirely mobile nodes that can establish communication in the absence of any fixed infrastructure. Envisioned applications of these networks include virtual classrooms, emergency relief operations, military tactical communications, sensor networks and community networking. Mobile ad hoc networking poses several new challenges in the design of network protocols. This thesis focuses on the routing problem. The main challenges in the design of a routing protocol for mobile ad hoc networks result from them having limited resources and there being frequent topological changes that occur unpredictably. Moreover, there is no fixed infrastructure that supports routing. The conventional routing protocols are not generally suitable for mobile ad hoc networks, as they cannot react quickly to the changing network topology, cause excessive communication and computation, or converge very slowly creating routing loops. In this thesis we propose two classes of routing schemes for mobile ad hoc networks. The first class is known as Limited Flooding Protocol. The protocol is fully reactive and does not require the computation of routing tables. It uses some basic principles of flooding, but reduces the communication overhead by restricting packet propagation through the network. Several variations of limited flooding are considered including deterministic, randomised and priority-based mechanisms. The main advantage of this protocol is that it can be used in networks with unpredictable topological changes and highly mobile nodes, since maintaining routing table at the intermediate nodes is not required. The second class of routing protocols is based on hierarchical clustering architecture and is intended for use in a relatively low mobility environment. The basic idea of this protocol is to partition the entire network into smaller units known as clusters and define routing mechanisms both within and between clusters using a hierarchical architecture. The main advantage of this architecture is reduction of storage requirements of routing information, communication overhead and computational overhead at each node. Discrete-event simulation is used for modelling and performance evaluation. Various options and variations of the protocols are examined in the…[Page 2 of abstract is missing.]
Detection and characterisation of vessels in retinal images.
(2015) Mapayi, Temitope.; Viriri, Serestina.; Tapamo, Jules-Raymond.
As retinopathies such as diabetic retinopathy (DR) and retinopathy of prematurity (ROP) continue to be the major causes of blindness globally, regular retinal examinations of patients can assist in the early detection of the retinopathies. The manual detection of retinal vessels is a very tedious and time consuming task as it requires about two hours to manually detect vessels in each retinal image. Automatic vessel segmentation has been helpful in achieving speed, improved diagnosis and progress monitoring of these diseases but has been challenging due to complexities such as the varying width of the retinal vessels from very large to very small, low contrast of thin vessels with respect to background and noise due to nonhomogeneous illumination in the retinal images. Although several supervised and unsupervised segmentation methods have been proposed in the literature, the segmentation of thinner vessels, connectivity loss of the vessels and time complexity remain the major challenges. In order to address these problems, this research work investigated di erent unsupervised segmentation approaches to be used in the robust detection of large and thin retinal vessels in a timely e cient manner. Firstly, this thesis conducted a study on the use of di erent global thresholding techniques combined with di erent pre-processing and post-processing techniques. Two histogram-based global thresholding techniques namely, Otsu and Isodata were able to detect large retinal vessels but fail to segment the thin vessels because these thin vessels have very low contrast and are di cult to distinguish from the background tissues using the histogram of the retinal images. Two new multi-scale approaches of computing global threshold based on inverse di erence moment and sum-entropy combined with phase congruence are investigated to improve the detection of vessels. One of the findings of this study is that the multi-scale approaches of computing global threshold combined with phase congruence based techniques improved on the detection of large vessels and some of the thin vessels. They, however, failed to maintain the width of the detected vessels. The reduction in the width of the detected large and thin vessels results in low sensitivity rates while relatively good accuracy rates were maintained. Another study on the use of fuzzy c-means and GLCM sum entropy combined on phase congruence for vessel segmentation showed that fuzzy c-means combined with phase congruence achieved a higher average accuracy rates of 0.9431 and 0.9346 but a longer running time of 27.1 seconds when compared with the multi-scale based sum entropy thresholding combined with phase congruence with the average accuracy rates of 0.9416 and 0.9318 with a running time of 10.3 seconds. The longer running time of the fuzzy c-means over the sum entropy thresholding is, however, attributed to the iterative nature of fuzzy c-means. When compared with the literature, both methods achieved considerable faster running time. This thesis investigated two novel local adaptive thresholding techniques for the segmentation of large and thin retinal vessels. The two novel local adaptive thresholding techniques applied two di erent Haralick texture features namely, local homogeneity and energy. Although these two texture features have been applied for supervised image segmentation in the literature, their novelty in this thesis lies in that they are applied using an unsupervised image segmentation approach. Each of these local adaptive thresholding techniques locally applies a multi-scale approach on each of the texture information considering the pixel of interest in relationship with its spacial neighbourhood to compute the local adaptive threshold. The localised multi-scale approach of computing the thresholds handled the challenge of the vessels' width variation. Experiments showed significant improvements in the average accuracy and average sensitivity rates of these techniques when compared with the previously discussed global thresholding methods and state of the art. The two novel local adaptive thresholding techniques achieved a higher reduction of false vessels around the border of the optic disc when compared with some of the previous techniques in the literature. These techniques also achieved a highly improved computational time of 1.9 to 3.9 seconds to segment the vessels in each retinal image when compared with the state of the art. Hence, these two novel local adaptive thresholding techniques are proposed for the segmentation of the vessels in the retinal images. This thesis further investigated the combination of di erence image and kmeans clustering technique for the segmentation of large and thin vessels in retinal images. The pre-processing phase computed a di erence image and k-means clustering technique was used for the vessel detection. While investigating this vessel segmentation method, this thesis established the need for a difference image that preserves the vessel details of the retinal image. Investigating the di erent low pass filters, median filter yielded the best di erence image required by k-means clustering for the segmentation of the retinal vessels. Experiments showed that the median filter based di erence images combined with k-means clustering technique achieved higher average accuracy and average sensitivity rates when compared with the previously discussed global thresholding methods and the state of the art. The median filter based di erence images combined with k-means clustering technique (that is, DIMDF) also achieved a higher reduction of false vessels around the border of the optic disc when compared with some previous techniques in the literature. These methods also achieved a highly improved computational time of 3.4 to 4 seconds when compared with the literature. Hence, the median filter based di erence images combined with k-means clustering technique are proposed for the segmentation of the vessels in retinal images. The characterisation of the detected vessels using tortuosity measure was also investigated in this research. Although several vessel tortuosity methods have been discussed in the literature, there is still need for an improved method that e ciently detects vessel tortuosity. The experimental study conducted in this research showed that the detection of the stationary points helps in detecting the change of direction and twists in the vessels. The combination of the vessel twist frequency obtained using the stationary points and distance metric for the computation of normalised and nonnormalised tortuosity index (TI) measure was investigated. Experimental results showed that the non-normalised TI measure had a stronger correlation with the expert's ground truth when compared with the distance metric and normalised TI measures. Hence, a non-normalised TI measure that combines the vessel twist frequency based on the stationary points and distance metric is proposed for the measurement of vessel tortuosity.
The enhanced best performance algorithm for global optimization with applications.
(2016) Chetty, Mervin.; Adewumi, Aderemi Oluyinka.
Abstract available in PDF file.
Exploration of ear biometrics with deep learning.
(2024) Booysens, Aimee Anne.; Viriri, Serestina.
Biometrics is the recognition of a human using biometric characteristics for identification, which may be physiological or behavioural. Numerous models have been proposed to distinguish biometric traits used in multiple applications, such as forensic investigations and security systems. With the COVID-19 pandemic, facial recognition systems failed due to users wearing masks; however, human ear recognition proved more suitable as it is visible. This thesis explores efficient deep learning-based models for accurate ear biometrics recognition. The ears were extracted and identified from 2D profiles and facial images, focusing on both left and right ears. With the numerous datasets used, with particular mention of BEAR, EarVN1.0, IIT, ITWE and AWE databases. Many machine learning techniques were explored, such as Naïve Bayes, Decision Tree, K-Nearest Neighbor, and innovative deep learning techniques: Transformer Network Architecture, Lightweight Deep Learning with Model Compression and EfficientNet. The experimental results showed that the Transformer Network achieved a high accuracy of 92.60% and 92.56% with epochs of 50 and 90, respectively. The proposed ReducedFireNet Model reduces the input size and increases computation time, but it detects more robust ear features. The EfficientNet variant B8 achieved a classification accuracy of 98.45%. The results achieved are more significant than those of other works, with the highest achieved being 98.00%. The overall results showed that deep learning models can improve ear biometrics recognition when both ears are computed.
Facial expression recognition and intensity estimation.
(2022) Ekundayo, Olufisayo Sunday.; Viriri, Serestina.
Facial Expression is one of the profound non-verbal channels through which human emotion state is inferred from the deformation or movement of face components when facial muscles are activated. Facial Expression Recognition (FER) is one of the relevant research fields in Computer Vision (CV) and Human-Computer Interraction (HCI). Its application is not limited to: robotics, game, medical, education, security and marketing. FER consists of a wealth of information. Categorising the information into primary emotion states only limit its performance. This thesis considers investigating an approach that simultaneously predicts the emotional state of facial expression images and the corresponding degree of intensity. The task also extends to resolving FER ambiguous nature and annotation inconsistencies with a label distribution learning method that considers correlation among data. We first proposed a multi-label approach for FER and its intensity estimation using advanced machine learning techniques. According to our findings, this approach has not been considered for emotion and intensity estimation in the field before. The approach used problem transformation to present FER as a multilabel task, such that every facial expression image has unique emotion information alongside the corresponding degree of intensity at which the emotion is displayed. A Convolutional Neural Network (CNN) with a sigmoid function at the final layer is the classifier for the model. The model termed ML-CNN (Multilabel Convolutional Neural Network) successfully achieve concurrent prediction of emotion and intensity estimation. ML-CNN prediction is challenged with overfitting and intraclass and interclass variations. We employ Visual Geometric Graphics-16 (VGG-16) pretrained network to resolve the overfitting challenge and the aggregation of island loss and binary cross-entropy loss to minimise the effect of intraclass and interclass variations. The enhanced ML-CNN model shows promising results and outstanding performance than other standard multilabel algorithms. Finally, we approach data annotation inconsistency and ambiguity in FER data using isomap manifold learning with Graph Convolutional Networks (GCN). The GCN uses the distance along the isomap manifold as the edge weight, which appropriately models the similarity between adjacent nodes for emotion predictions. The proposed method produces a promising result in comparison with the state-of-the-art methods.
Forest image classification based on deep learning and ontologies.
(2024) Kwenda, Clopas.; Gwetu, Mandlenkosi Victor.; Fonou-Dombeu, Jean Vincent.
Forests contribute abundantly to nature’s natural resources and they significantly contribute to a wide range of environmental, socio-cultural, and economic benefits. Classifications of forest vegetation offer a practical method for categorising information about patterns of forest vegetation. This information is required to successfully plan for land use, map landscapes, and preserve natural habitats. Remote sensing technology has provided high spatio-temporal resolution images with many spectral bands that make conducting research in forestry easy. In that regard, artificial intelligence technologies assess forest damage. The field of remote sensing research is constantly adapting to leverage newly developed computational algorithms and increased computing power. Both the theory and the practice of remote sensing have significantly changed as a result of recent technological advancements, such as the creation of new sensors and improvements in data accessibility. Data-driven methods, including supervised classifiers (such as Random Forests) and deep learning classifiers, are gaining much importance in processing big earth observation data due to their accuracy in creating observable images. Though deep learning models produce satisfactory results, researchers find it difficult to understand how they make predictions because they are regarded as black-box in nature, owing to their complicated network structures. However, when inductive inference from data learning is taken into consideration, data-driven methods are less efficient in working with symbolic information. In data-driven techniques, the specialized knowledge that environmental scientists use to evaluate images obtained through remote sensing is typically disregarded. This limitation presents a significant obstacle for end users of Earth Observation applications who are accustomed to working with symbolic information, such as ecologists, agronomists, and other related professionals. This study advocates for the incorporation of ontologies in forest image classification owing to their ability in representing domain expert knowledge. The future of remote sensing science should be supported by knowledge representation techniques such as ontologies. The study presents a methodological framework that integrates deep learning techniques and ontologies with the aim of enhancing domain expert confidence as well as increasing the accuracy of forest image classification. In addressing this challenge, this study followed the following systematic steps (i) A critical review of existing methods for forest image classification (ii) A critical analysis of appropriate methods for forest image classification (iii) Development of the state-of-the-art model for forest image segmentation (iv) Design of a hybrid model of deep learning and machine learning model for forest image classification (v) A state-of-the-art ontological framework for forest image classification. The ontological framework was flexible to capture the expression of the domain expert knowledge. The ontological state-of-the-art model performed well as it achieved a classification accuracy of 96%, with a Root Mean Square Error of 0.532. The model can also be used in the fruit industry and supermarkets to classify fruits into their respective categories. It can also be potentially used to classify trees with respect to their species. As a way of enhancing confidence in deep learning models by domain experts, the study recommended the adoption of explainable artificial intelligence (XAI) methods because they unpack the process by which deep learning models reach their decision. The study also recommended the adoption of high-resolution networks (HRNets) as an alternative to traditional deep learning models, because they can convert low-resolution representation to high-resolution and have efficient block structures developed according to new standards and they are excellent at being used for feature extraction.
Formalisms for agents reasoning with stochastic actions and perceptions.
(2014) Rens, Gavin Brian.; Meyer, Thomas Andreas.; Lakemeyer, Gerhard.
The thesis reports on the development of a sequence of logics (formal languages based on mathematical logic) to deal with a class of uncertainty that agents may encounter. More accurately, the logics are meant to be used for allowing robots or software agents to reason about the uncertainty they have about the effects of their actions and the noisiness of their observations. The approach is to take the well-established formalism called the partially observable Markov decision process (POMDP) as an underlying formalism and then design a modal logic based on POMDP theory to allow an agent to reason with a knowledge-base (including knowledge about the uncertainties). First, three logics are designed, each one adding one or more important features for reasoning in the class of domains of interest (i.e., domains where stochastic action and sensing are considered). The final logic, called the Stochastic Decision Logic (SDL) combines the three logics into a coherent formalism, adding three important notions for reasoning about stochastic decision-theoretic domains: (i) representation of and reasoning about degrees of belief in a statement, given stochastic knowledge, (ii) representation of and reasoning about the expected future rewards of a sequence of actions and (iii) the progression or update of an agent’s epistemic, stochastic knowledge. For all the logics developed in this thesis, entailment is defined, that is, whether a sentence logically follows from a knowledge-base. Decision procedures for determining entailment are developed, and they are all proved sound, complete and terminating. The decision procedures all employ tableau calculi to deal with the traditional logical aspects, and systems of equations and inequalities to deal with the probabilistic aspects. Besides promoting the compact representation of POMDP models, and the power that logic brings to the automation of reasoning, the Stochastic Decision Logic is novel and significant in that it allows the agent to determine whether or not a set of sentences is entailed by an arbitrarily precise specification of a POMDP model, where this is not possible with standard POMDPs. The research conducted for this thesis has resulted in several publications and has been presented at several workshops, symposia and conferences.
Hierarchical age estimation using enhanced facial features.
(2018) Angulu, Raphael.; Tapamo, Jules-Raymond.; Adewumi, Aderemi Oluyinka.
Ageing is a stochastic, inevitable and uncontrollable process that constantly affect shape, texture and general appearance of the human face. Humans can easily determine ones’ gender, identity and ethnicity with highest accuracy as compared to age. This makes development of automatic age estimation techniques that surpass human performance an attractive yet challenging task. Automatic age estimation requires extraction of robust and reliable age discriminative features. Local binary patterns (LBP) sensitivity to noise makes it insufficiently reliable in capturing age discriminative features. Although local ternary patterns (LTP) is insensitive to noise, it uses a single static threshold for all images regardless of varied image conditions. Local directional patterns (LDP) uses k directional responses to encode image gradient and disregards not only central pixel in the local neighborhood but also 8 􀀀 k directional responses. Every pixel in an image carry subtle information. Discarding 8 􀀀 k directional responses lead to lose of discriminative texture features. This study proposes two variations of LDP operator for texture extraction. Significantorientation response LDP (SOR-LDP) encodes image gradient by grouping eight directional responses into four pairs. Each pair represents orientation of an edge with respect to central reference pixel. Values in each pair are compared and the bit corresponding to the maximum value in the pair is set to 1 while the other is set to 0. The resultant binary code is converted to decimal and assigned to the central pixel as its’ SOR-LDP code. Texture features are contained in the histogram of SOR-LDP encoded image. Local ternary directional patterns (LTDP) first gets the difference between neighboring pixels and central pixel in 3 3 image region. These differential values are convolved with Kirsch edge detectors to obtain directional responses. These responses are normalized and used as probability of an edge occurring towards a respective direction. An adaptive threshold is applied to derive LTDP code. The LTDP code is split into its positive and negative LTDP codes. Histograms of negative and positive LTDP encoded images are concatenated to obtain texture feature. Regardless of there being evidence of spatial frequency processing in primary visual cortex, biologically inspired features (BIF) that model visual cortex uses only scale and orientation selectivity in feature extraction. Furthermore, these BIF are extracted using holistic (global) pooling across scale and orientations leading to lose of substantive information. This study proposes multi-frequency BIF (MF-BIF) where frequency selectivity is introduced in BIF modelling. Local statistical BIF (LS-BIF) uses local pooling within scale, orientation and frequency in n n region for BIF extraction. Using Leave-one-person-out (LOPO) validation protocol, this study investigated performance of proposed feature extractors in age estimation in a hierarchical way by performing age-group classification using Multi-layer Perceptron (MLP) followed by within age-group exact age regression using support vector regression (SVR). Mean absolute error (MAE) and cumulative score (CS) were used to evaluate performance of proposed face descriptors. Experimental results on FG-NET ageing dataset show that SOR-LDP, LTDP, MF-BIF and LS-BIF outperform state-of-the-art feature descriptors in age estimation. Experimental results show that performing gender discrimination before age-group and age estimation further improves age estimation accuracies. Shape, appearance, wrinkle and texture features are simultaneously extracted by visual system in primates for the brain to process and understand an image or a scene. However, age estimation systems in the literature use a single feature for age estimation. A single feature is not sufficient enough to capture subtle age discriminative traits due to stochastic and personalized nature of ageing. This study propose fusion of different facial features to enhance their discriminative power. Experimental results show that fusing shape, texture, wrinkle and appearance result into robust age discriminative features that achieve lower MAE compared to single feature performance.
Improved roach-based algorithms for global optimization problems.
(2014) Obagbuwa, Ibidun Christiana.; Adewumi, Aderemi Oluyinka.
Optimization of systems plays an important role in various fields including mathematics, economics, engineering and life sciences. A lot of real world optimization problems exist across field of endeavours such as engineering design, space planning, networking, data analysis, logistic management, financial planning, risk management, and a host of others. These problems are constantly increasing in size and complexity, necessitating the need for improved techniques. Many conventional approaches have failed to solve complex problems effectively due to increasingly large solution space. This has led to the development of evolutionary algorithms that draw inspiration from the process of natural evolution. It is believed that nature provides inspirations that can lead to innovative models or techniques for solving complex optimization problems. Among the class of paradigm based on this inspiration is Swarm Intelligence (SI). SI is one of the recent developments of evolutionary computation. A SI paradigm is comprised of algorithms inspired by the social behaviour of animals and insects. SI-based algorithms have attracted interest, gained popularity and attention because of their flexibility and versatility. SIbased algorithms have been found to be efficient in solving real world optimization problems. Examples of SI algorithms include Ant Colony Optimization (ACO) inspired by the pheromone trail-following behaviour of ant species; Particle Swarm Optimization (PSO) inspired by flocking and swarming behaviour of insects and animals; and Bee Colony Optimization (BCO) inspired by bees’ food foraging. Recent emerging techniques in SI includes Roach-based Algorithms (RBA) motivated by cockroaches social behaviour. Two recently introduced RBA algorithms are Roach Infestation Optimization (RIO) and Cockroach Swarm Optimization (CSO) which have been applied to some optimization problems to achieve competitive results when compared to PSO. This study is motivated by the promising results of RBA, which have shown that the algorithms have potentials to be efficient tools for solving optimization problems. Extensive studies of existing RBA were carried out in this work revealing the shortcomings such as slow convergence and entrapment in local minima. The aim of this study is to overcome the identified drawbacks. We investigate RBA variants that are introduced in this work by introducing parameters such as constriction factor and sigmoid function that have proved effective for similar evolutionary algorithms in the literature. In addition components such as vigilance, cannibalism and hunger are incorporated into existing RBAs. These components are constructed by the use of some known techniques such as simple Euler, partial differential equation, crossover and mutation methods to speed up convergence and enhance the stability, exploitation and exploration of RBA. Specifically, a stochastic constriction factor was introduced to the existing CSO algorithm to improve its performance and enhance its ability to solve optimization problems involving thousands of variables. A CSO algorithm that was originally designed with three components namely chase-swarming, dispersion and ruthlessness is extended in this work with hunger component to improve its searching ability and diversity. Also, predator-prey evolution using crossover and mutation techniques were introduced into the CSO algorithm to create an adaptive search in each iteration thereby making the algorithm more efficient. In creating a discrete version of a CSO algorithm that can be used to evaluate optimization problems with any discrete range value, we introduced the sigmoid function. Furthermore, a dynamic step-size adaptation with simple Euler method was introduced to the existing RIO algorithm enhancing swarm stability and improving local and global searching abilities. The existing RIO model was also re-designed with the inclusion of vigilance and cannibalism components. The improved RBA were tested on established global optimization benchmark problems and results obtained compared with those from the literature. The improved RBA introduced in this work show better improvements over existing ones.
Intelligent instance selection techniques for support vector machine speed optimization with application to e-fraud detection.
(2017) Akinyelu, Ayobami Andronicus.; Adewumi, Aderemi Oluyinka.
Decision-making is a very important aspect of many businesses. There are grievous penalties involved in wrong decisions, including financial loss, damage of company reputation and reduction in company productivity. Hence, it is of dire importance that managers make the right decisions. Machine Learning (ML) simplifies the process of decision making: it helps to discover useful patterns from historical data, which can be used for meaningful decision-making. The ability to make strategic and meaningful decisions is dependent on the reliability of data. Currently, many organizations are overwhelmed with vast amounts of data, and unfortunately, ML algorithms cannot effectively handle large datasets. This thesis therefore proposes seven filter-based and five wrapper-based intelligent instance selection techniques for optimizing the speed and predictive accuracy of ML algorithms, with a particular focus on Support Vector Machine (SVM). Also, this thesis proposes a novel fitness function for instance selection. The primary difference between the filter-based and wrapper-based technique is in their method of selection. The filter-based techniques utilizes the proposed fitness function for selection, while the wrapper-based technique utilizes SVM algorithm for selection. The proposed techniques are obtained by fusing SVM algorithm with the following Nature Inspired algorithms: flower pollination algorithm, social spider algorithm, firefly algorithm, cuckoo search algorithm and bat algorithm. Also, two of the filter-based techniques are boundary detection algorithms, inspired by edge detection in image processing and edge selection in ant colony optimization. Two different sets of experiments were performed in order to evaluate the performance of the proposed techniques (wrapper-based and filter-based). All experiments were performed on four datasets containing three popular e-fraud types: credit card fraud, email spam and phishing email. In addition, experiments were performed on 20 datasets provided by the well-known UCI data repository. The results show that the proposed filter-based techniques excellently improved SVM training speed in 100% (24 out of 24) of the datasets used for evaluation, without significantly affecting SVM classification quality. Moreover, experimental results also show that the wrapper-based techniques consistently improved SVM predictive accuracy in 78% (18 out of 23) of the datasets used for evaluation and simultaneously improved SVM training speed in all cases. Furthermore, two different statistical tests were conducted to further validate the credibility of the results: Freidman’s test and Holm’s post-hoc test. The statistical test results reveal that the proposed filter-based and wrapper-based techniques are significantly faster, compared to standard SVM and some existing instance selection techniques, in all cases. Moreover, statistical test results also reveal that Cuckoo Search Instance Selection Algorithm outperform all the proposed techniques, in terms of speed. Overall, the proposed techniques have proven to be fast and accurate ML-based e-fraud detection techniques, with improved training speed, predictive accuracy and storage reduction. In real life application, such as video surveillance and intrusion detection systems, that require a classifier to be trained very quickly for speedy classification of new target concepts, the filter-based techniques provide the best solutions; while the wrapper-based techniques are better suited for applications, such as email filters, that are very sensitive to slight changes in predictive accuracy.
The investigation into an algorithm based on wavelet basis functions for the spatial and frequency decomposition of arbitrary signals.
(1994) Goldstein, Hilton.; Sartori-Angus, Alan G.
The research was directed toward the viability of an O(n) algorithm which could decompose an arbitrary signal (sound, vibration etc.) into its time-frequency space. The well known Fourier Transform uses sine and cosine functions (having infinite support on t) as orthonormal basis functions to decompose a signal i(t) in the time domain to F(w) in the frequency . domain, where the Fourier coefficients F(w) are the contributions of each frequency in the original signal. Due to the non-local support of these basis functions, a signal containing a sharp localised transient does not have localised coefficients, but rather coefficients that decay slowly. Another problem is that the coefficients F(w) do not convey any time information. The windowed Fourier Transform, or short-time Fourier Transform, does attempt to resolve the latter, but has had limited success. Wavelets are basis functions, usually mutually orthonormal, having finite support in t and are therefore spatially local. Using non-orthogonal wavelets, the Dominant Scale Transform (DST) designed by the author, decomposes a signal into its approximate time-frequency space. The associated Dominant Scale Algorithm (DSA) has O(n) complexity and is integer-based. These two characteristics make the DSA extremely efficient. The thesis also investigates the problem of converting a music signal into it's equivalent music score. The old problem of speech recognition is also examined. The results obtained from the DST are shown to be consistent with those of other authors who have utilised other methods. The resulting DST coefficients are shown to render the DST particularly useful in speech segmentation (silence regions, voiced speech regions, and frication). Moreover, the Spectrogram Dominant Scale Transform (SDST), formulated from the DST, was shown to approximate the Fourier coefficients over fixed time intervals within vowel regions of human speech.
An investigation into the use of genetic programming for the induction of novice procedural programming solution algorithms in intelligent programming tutors.
(2004) Pillay, Nelishia.; Sartori-Angus, Alan G.
Intelligent programming tutors have proven to be an economically viable and effective means of assisting novice programmers overcome learning difficulties. However, the large-scale use of intelligent programming tutors has been impeded by the high developmental costs associated with building intelligent programming tutors. The research presented in this thesis forms part of a larger initiative aimed at reducing these costs by building a generic architecture for the development of intelligent programming tutors. One of the facilities that must be provided by the generic architecture is the automatic generation of solutions to programming problems. The study presented in the thesis examines the use of genetic programming as means of inducing solution algorithms to novice programming problems. The scope of the thesis is limited to novice procedural programming paradigm problems requiring the use of arithmetic, string manipulation, conditional, iterative and recursive programming structures. The methodology employed in the study is proof-by-demonstration. A genetic programming system for the induction of novice procedural solution algorithms was implemented and tested on randomly chosen novice procedural programming problems. The study has identified the standard and advanced genetic programming features needed for the successful generation of novice procedural solution algorithms. The outcomes of this study include the derivation of an internal representation language for representing procedural solution algorithms and a high-level programming problem specification format for describing procedural problems, in the generic architecture. One of the limitations of genetic programming is its susceptibility to converge prematurely to local optima and not find a solution in some cases. The study has identified fitness function biases against certain structural components that are needed to find a solution, as an additional cause of premature convergence in this domain. It presents an iterative structure-based algorithm as a solution to this problem. This thesis has contributed to both the fields of genetic programming and intelligent programming tutors. While genetic programming has been successfully implemented in various domains, it is usually applied to a single problem within that domain. In this study the genetic programming system must be capable of solving a number of different programming problems in different application domains. In addition to this, the study has also identified a means of overcoming premature convergence caused by fitness function biases in a genetic programming system for the induction of novice procedural programming algorithms. Furthermore, although a number of studies have addressed the student modelling and pedagogical aspects of intelligent programming tutors, none have examined the automatic generation of problem solutions as a means of reducing developmental costs. Finally, this study has contributed to the ongoing research being conducted by the artificial intelligence in education community, to test the effectiveness of using machine learning techniques in the development of different aspects of intelligent tutoring systems.
A knowledge-based system for automated discovery of ecological interactions in flower-visiting data.
(2017) Coetzer, Willem Gabriël.; Moodley, Deshendran.; Gerber, Aurona Jacoba.
Studies on the community ecology of flower-visiting insects, which can be inferred to pollinate flowers, are important in agriculture and nature conservation. Many scientific observations of flower-visiting insects are associated with digitized records of insect specimens preserved in natural history collections. Specimen annotations include heterogeneous and incomplete, in situ field documentation of ecologically significant relationships between individual organisms (i.e. insects and plants), which are nevertheless potentially valuable. A wealth of unrepresented biodiversity and ecological knowledge can be unlocked from such detailed data by augmenting the data with expert knowledge encoded in knowledge models. An analysis of the knowledge representation requirements of flower-visiting community ecologists is presented, as well as an implementation and evaluation of a prototype knowledge-based system for automated semantic enrichment, semantic mediation and interpretation of flower-visiting data. A novel component of the system is a semantic architecture which incorporates knowledge models validated by experts. The system combines ontologies and a Bayesian network to enrich, integrate and interpret flower- visiting data, specifically to discover ecological interactions in the data. The system’s effectiveness, to acquire and represent expert knowledge and simulate the inferencing ability of expert flower-visiting ecologists, is evaluated and discussed. The knowledge-based system will allow a novice ecologist to use standardised semantics to construct interaction networks automatically and objectively. This could be useful, inter alia, when comparing interaction networks for different periods of time at the same place or different places at the same time. While the system architecture encompasses three levels of biological organization, data provenance can be traced back to occurrences of individual organisms preserved as evidence in natural history collections. The potential impact of the semantic architecture could be significant in the field of biodiversity and ecosystem informatics because ecological interactions are important in applied ecological studies, e.g. in freshwater biomonitoring or animal migration.
Leaf recognition for accurate plant classification.
(2017) Kala, Jules Raymond.; Viriri, Serestina.; Moodley, Deshendran.
Plants are the most important living organisms on our planet because they are sources of energy and protect our planet against global warming. Botanists were the first scientist to design techniques for plant species recognition using leaves. Although many techniques for plant recognition using leaf images have been proposed in the literature, the precision and the quality of feature descriptors for shape, texture, and color remain the major challenges. This thesis investigates the precision of geometric shape features extraction and improved the determination of the Minimum Bounding Rectangle (MBR). The comparison of the proposed improved MBR determination method to Chaudhuri's method is performed using Mean Absolute Error (MAE) generated by each method on each edge point of the MBR. On the top left point of the determined MBR, Chaudhuri's method has the MAE value of 26.37 and the proposed method has the MAE value of 8.14. This thesis also investigates the use of the Convexity Measure of Polygons for the characterization of the degree of convexity of a given leaf shape. Promising results are obtained when using the Convexity Measure of Polygons combined with other geometric features to characterize leave images, and a classification rate of 92% was obtained with a Multilayer Perceptron Neural Network classifier. After observing the limitations of the Convexity Measure of Polygons, a new shape feature called Convexity Moments of Polygons is presented in this thesis. This new feature has the invariant properties of the Convexity Measure of Polygons, but is more precise because it uses more than one value to characterize the degree of convexity of a given shape. Promising results are obtained when using the Convexity Moments of Polygons combined with other geometric features to characterize the leaf images and a classification rate of 95% was obtained with the Multilayer Perceptron Neural Network classifier. Leaf boundaries carry valuable information that can be used to distinguish between plant species. In this thesis, a new boundary-based shape characterization method called Sinuosity Coefficients is proposed. This method has been used in many fields of science like Geography to describe rivers meandering. The Sinuosity Coefficients is scale and translation invariant. Promising results are obtained when using Sinuosity Coefficients combined with other geometric features to characterize the leaf images, a classification rate of 80% was obtained with the Multilayer Perceptron Neural Network classifier. Finally, this thesis implements a model for plant classification using leaf images, where an input leaf image is described using the Convexity Moments, the Sinuosity Coefficients and the geometric features to generate a feature vector for the recognition of plant species using a Radial Basis Neural Network. With the model designed and implemented the overall classification rate of 97% was obtained.

Browse

Browsing Doctoral Degrees (Computer Science) by Title