21793 research outputs found
Sort by
Improving Visual Processing through Feature Descriptor Enhancement and Depth Upsampling
No full textVisual processing is a core aspect of modern computer vision, supporting applications such as object recognition, tracking, and 3D reconstruction. However, challenges like occlusions, appearance variations, and inconsistencies between RGB and depth data continue to hinder reliable scene interpretation. Traditional feature descriptors often lack adaptability in complex environments, and depth maps acquired from low-cost sensors typically exhibit low resolution, noise, and missing values. These limitations obstruct the accurate extraction of spatial and contextual information required for robust scene analysis. To address these issues, the overall objective of this thesis is to advance visual processing through adaptive, feature-driven techniques that leverage the spatial and contextual richness of color and depth data, thereby contributing to more reliable interpretation of complex visual scenes. Aligned with this objective, the thesis is structured around two main parts aimed at achieving more accurate and context-aware scene understanding.
The first part of the thesis is focused on the design of two advanced feature descriptors and an object tracking scheme. The first descriptor, r-spatiogram, captures spatial, color, and texture information within image regions to provide detail-preserving, context-aware representation. The second descriptor, adaptive multi-scale (AMS), improves adaptability by employing strategies suited to diverse visual environments. In continuation of this part, a novel object tracking
iv
framework is introduced that utilizes color and depth information to address common challenges such as occlusions and target-background similarity. This framework initially operates independently and is then extended by incorporating each proposed descriptor to evaluate their impact on tracking accuracy and robustness in dynamic environments.
The second part of the thesis is concerned with a novel depth upsampling scheme that improves the quality of low-resolution depth maps. It employs a joint local–nonlocal framework guided by an adaptive bandwidth mechanism that dynamically adjusts the influence of neighboring pixels. A distance-based patch similarity map is introduced to support this adaptation. Two similarity strategies are explored, one using a standard metric, and the other incorporating the AMS descriptor for capturing more complex structural relationships.
Extensive experiments are conducted on multiple benchmark datasets to validate the effectiveness of the proposed schemes
Optimal Cybersecurity of Cyber-Physical Systems Using Game-Theory Strategies
No full textIn this thesis, the problem of optimal cybersecurity of Cyber-Physical Systems (CPS) will be addressed. In the first topic, optimal cybersecurity of CPS based on game theory method will be studied. In this problem, cybersecurity of CPS is investigated by defining a new game between attacker and defender in the case that
there is a passive detection mechanism on the Command-and-Control (C&C) side.
By considering the defender’s fixed strategy, a game theory-based optimal attack in the viewpoint of attacker will be designed. The proposed cyber-attack strategy can deteriorate the performance of the system and steer the system to the desired performance of the attacker and remains stealthy. Towards these ends, regarding
the general structure of a CPS, two cooperating attackers on input channels and measurement channels are considered to inject their signals according to the strategy of each other.
According to the defined objectives of each attacker, an appropriate cost function is defined for each attacker that depends on their strategies as well. Defining an optimization problem results in some coupled Hamilton Jacobi Belman (HJB) equations. To solve these coupled complex equations, some machine learning methods
will be applied to reach an optimal attack strategy. It will be shown that the passive detection mechanism cannot detect the proposed optimal attack strategy.
Furthermore, game theory-based robust optimal attack design in the viewpoint of attacker for defender’s fixed strategy detector in the presence of disturbances will be addressed to verify the detectability condition of the optimal cyber-attacks. To
address the effect of the disturbance in the system for the defined game, a hierarchical game that is the combination of zero-sum and Stackelberg game is proposed.
In the second problem, an optimal secure estimation and resilient control for linear CPS by taking advantages of the game theory method will be presented. To have the capability of the optimal secure estimation, we use an auxiliary system to provide the redundancy information on the measurement of the main plant. The
information of the auxiliary system with a specific dynamic will be added to the output of the system. As a result, we have a new set of information with redundancy to be transferred on the communication channels. To make it difficult for the attacker
to find the channels that are communicating information, a Moving Target Defense (MTD) block is utilized on the plant side before sending the information.
On the other hand, to have the same capability on the input channels to provide a resilient strategy, an auxiliary system will be used on the C&C side in parallel with the main controller to provide redundancies in the input signals. By adopting the same strategy on the measurement channels, i.e. using Moving Target Defense (MTD) approach, attacker cannot identify the channels that are communicating information.
Finally, in the third sub-problem of this chapter, we will consider the case that there is the possibility of the actuator fault as well as cyber-attacks. The problem will be solved based on game theory method by taking advantages of reinforcement learning
method
Log-Based Anomaly Detection: Comparative Study of Real-World System Logs using Machine Learning And Deep Learning Approaches
No full textThe reliability and security of today’s smart and autonomous systems increasingly rely on effective anomaly detection capabilities. Logs generated by intelligent devices during runtime offer valuable insights for system monitoring and troubleshooting. Nonetheless, the enormous quantity and complexity of these logs render manual anomaly inspection impractical and error-prone. To address this, various automated log-based anomaly detection methods have been developed. However, many of these approaches are evaluated in controlled environments with publicly available datasets, which differ significantly from the noisy, unstructured, and unlabeled logs encountered in industrial settings. This thesis explores and adapts existing machine learning and deep learning techniques for anomaly detection in real-world system logs produced by an intelligent autonomous display device. Initially, we conduct a comparative analysis of machine learning and deep learning methods using a small manually labeled dataset to evaluate the detection accuracy and computational efficiency. Our results highlight the most suitable approaches for enabling proactive maintenance and enhancing system reliability. Expanding on this, we evaluate advanced deep learning methods across weakly supervised, semi-supervised, and unsupervised learning paradigms, using heuristically labeled logs and benchmark them against fully supervised baselines to examine the trade-offs between label dependency, detection performance, and industrial applicability. Finally, we propose a systematic approach for managing unlabeled and noisy log data, providing practical guidelines for selecting suitable learning strategies based on label availability, data quality, and real-world constraints. The findings of this work provide valuable insights for the implementation of scalable, accurate, and robust log-based anomaly detection in industrial IoT environments
Health-Related Communication in Adolescents and Young Adults with Sickle Cell Disease
No full textBackground and Aims: Effective health-related communication is essential for pain and disease management but is poorly understood among adolescents and young adults (AYAs) with sickle cell disease (SCD). The current study aimed to: (1) Characterize the quality and prevalence of communication difficulties among AYAs with SCD; (2) Identify sociodemographic/disease-related factors associated with communication difficulties; and (3) Examine associations between pain, pain interference, and communication difficulties.
Method: AYAs with SCD (N=416; mean [M] age=18.6 years, range=13.2-25.0, 50% female, 99.8% Black; n=136 adolescents; n=280 young adults) enrolled in the Sickle Cell Clinical Research and Intervention Program completed the PedsQL-SCD, which contains subscales measuring self-advocacy (Communication I), eliciting understanding (Communication II), pain (Pain and Hurt), and pain interference (Pain Impact). Clinical data were extracted from medical records. Analyses included descriptive statistics, multivariate regressions (adjusted for age, sex, and genotype), with false discovery rate corrections.
Results: AYAs reported moderate-to-good quality communication (M=20.6 for self-advocacy; M=37.2 for eliciting understanding). However, 26.4% of AYAs reported feeling misunderstood about their pain and 22.3% reported feeling misunderstood about their disease. Female sex, greater economic hardship, older age, and Hb SS/Sβ⁰ genotype were associated with greater communication difficulties (all ps<0.05). Elevated pain and pain interference were also associated with communication difficulties (p<0.001).
Conclusion: These findings improve our understanding of health-related communication among AYAs with SCD and highlight factors associated with communication difficulties. Tailored interventions are needed to improve care and health outcomes among this population
Digital Financial Services and Cybersecurity: Barriers to Financial Inclusion in Developing Countries
No full textAs digital financial services grow across developing economies, they promise to close
long-standing gaps in financial inclusion. Yet, the rise of cyber threats and low digital
trust pose new challenges that can discourage adoption, especially among vulnerable
populations. This study examines how cybersecurity infrastructure and digital access
influence mobile money usage in 35 developing countries from 2018 to 2023. Using
panel data regression techniques, the results show the importance of building secure,
accessible, and user-trusted financial systems to drive inclusive growth in the digital
age
Safety and Operations of Autonomous Traffic at Highway Bottlenecks
No full textA reduction of the available travel lane, commonly referred to as a lane drop, may occur on freeways due to road design, incidents or road maintenance. Lane drops lead to merging and mandatory lane-changing, resulting in various traffic problems, including delays, congestion and safety risks. In mixed traffic environments, the interactions between autonomous vehicles and human-driven vehicles add complexity and alter traffic dynamics in uncertain ways. Understanding the performance of autonomous vehicles is essential for planning and developing a control framework. This study investigates AV performance at a lane-drop bottleneck under varying traffic demands and AV penetration rates, and explores the sensitivity of car-following and lane-changing behavioural parameters. Using PTV VISSIM-COM for microsimulation, three AV driving logics (i.e. cautious, normal, and aggressive) were modelled across four traffic demand levels. Safety performance was assessed using the Surrogate Safety Assessment Model (SSAM) based on surrogate conflicts indicators such as Time-to-Collision (TTC) and Post-Encroachment Time (PET). The results show trade-offs between safety and efficiency across driving logics. Cautious AVs enhance safety and flow stability at low penetration rates but lead to increased delays as penetration rises. Aggressive AVs reduce delays at high penetration rates but increase risk due to higher speed and more changeable behaviour. Normal AVs provide balanced performance across most conditions, particularly in moderate penetration scenarios. The findings emphasize the need for adaptive AV behaviour strategies that respond to real-time traffic composition, AV share and roadway complexity which is a key to achieving safe and efficient traffic systems
In the World, Not of It: Discourses of Toxicity in Conservative Christian Media
No full textAs the political power of the religious right mounts, so too does the importance of studying how conservative Christians shape their understanding of themselves in relation to the world around them. ‘Toxicity’ as a chemical, biological, and social phenomena is integrated within many contemporary conservative panics over what is and is not normal and what should and should not be tolerated. Drawing on medical humanities, religious media studies, disability studies, and feminist studies, this thesis approaches conservative Christian discourses of toxicity through a multidisciplinary discourse analysis. My research reaches across denominations and mediums to show the breadth of these discourses and their contemporary and historical implications. What I find is that discourses of toxicity are primarily employed in conservative Christian media to wrestle with issues that challenge this group’s political, spiritual, and religious assumptions about what (and who) does and does not belong. I integrate theoretical analysis throughout to illuminate how discourses of toxicity deal with the self/other divide and conclude by considering how these discourses relate to other concepts such as sin and abjection
From Industrial Park to Lifestyle Park: Shifting Narratives of Montreal’s Lachine Canal
No full textOnce Canada’s main manufacturing center, Montreal’s Lachine Canal supported more than six hundred factories from the mid-19th to the mid-20th century, becoming an incredibly diversified “industrial park” (Desloges & Gelly, 2002). Following deindustrialization in the 1970s and 1980s, its surrounding neighborhoods experienced economic decline. In the early 2000s, Parks Canada reopened the canal as a recreational corridor. This sparked a wave of condo development and the conversion of old industrial spaces into luxurious lofts and business hubs. These new spaces offer young professionals a curated lifestyle—one associated with living in a park, close to water, and in a heritage-conscious environment. Amid this rapid transformation, conflicting visions of the canal have emerged. Developers, government actors, and long-time residents assign different meanings to its landscape. Some emphasize environmental renewal and economic growth, while others fight to preserve working-class memory and public access. Acknowledging these tensions, this thesis explores how competing narratives of the Lachine Canal coexist within a rapidly gentrifying landscape. It focuses on how different meanings shaped by heritage, green aesthetics, and the symbolic significance of water are expressed and contested. Drawing on six months of ethnographic fieldwork and interviews—with municipal actors involved in the canal’s transformation, as well as with new and old residents who frequent the space—I show how these divergent meanings play out in everyday practices and interactions. This research also examines the role of material traces in remembrance, particularly of the canal’s industrial past. A key element of Montreal’s culture, the canal offers insight into the history, politics, heritage, power dynamics, and social life of the South-West borough. The project presents a nuanced account of how urban landscapes become contested arenas where the past, present, and future are negotiated
Spatial Variability of POC in Surface Waters of the St. Lawrence Estuary and Gulf: A Molecular and Bulk Analysis
No full textSuspended particulate matter (SPM) in aquatic systems is comprised of many inorganic and organic components with much of the organic matter not characterized. This study characterized and investigated the spatial variability of particulate organic carbon (POC) in the surface waters of the St. Lawrence Estuary and Gulf to discern the contributions of terrestrial and marine organic matter (OM). Using bulk elemental and isotopic analyses alongside lipid molecular biomarkers, specifically hydrocarbons and fatty acids, we characterized POC from surface water samples collected at 19 stations spanning eight scientific missions between 2003 and 2023. This comprehensive dataset is the first to incorporate both molecular and bulk analyses of POC in the surface waters of this system providing a baseline understanding of OM composition and source contributions in this dynamic system. Our findings reveal distinct spatial patterns in OM composition, with terrestrially derived OM dominating the Upper St. Lawrence Estuary (ULSE) and a gradual shift towards marine-derived OM as distance from Quebec City increases i.e., downriver. High molecular weight (HMW) n-alkanes (C27, C29, and C31) are prevalent in terrestrially influenced stations while low molecular weight (LMW) n-alkanes (C15, C17, and C19) dominate marine stations. Stable carbon isotopes (δ13C) and C/N ratios also reflect the transition from depleted to more enriched δ13C values along the gradient.
A multivariate approach was used to identify spatial variability using principal component analysis (PCA), broken stick analysis, and SIMPER. Using these techniques, we attempted to identify the primary drivers of OM composition across the continuum. Salinity, n-alkane proxies, and distance from Quebec emerged as key factors influencing OM distribution. Our results suggest that OM composition in the SLEG (St. Lawrence Estuary and Gulf) is controlled by both hydrodynamic processes and terrestrial-marine interactions. This study establishes a critical baseline for understanding the sources and spatial variability of OM in the SLEG which contribute valuable information into the biogeochemical processes that shape this important system. These findings emphasize the need for continued monitoring to evaluate future changes driven by natural and anthropogenic changes as well as climate change
AI Goes to Hollywood: Artificial Intelligence in the 2023 WGA and SAG-AFTRA Strikes
No full textIn 2023, the Writers Guild of America (WGA) and Screen Actors Guild-American Federation of Radio and Television Artists (SAG-AFTRA) went on strike at the same time, shutting down Hollywood film and television production for seven months. The potential role of artificial intelligence (AI) in the creative labour process emerged as a key theme in negotiations for both unions. This thesis asks: How did AI become a major issue in the 2023 dual strikes and what
does this reveal about Hollywood labour in the early 2020s? By taking a historical critical media industries framework, this project places the 2023 strikes in the context of a century of organized
Hollywood labour struggles, including previous strikes organized in response to emerging technology, as well as the distribution shift to online video streaming in the 2010s. Using critical discourse analysis of entertainment news coverage of the strikes in Deadline, The Hollywood Reporter and The New York Times, this thesis examines union members’ statements about AI to theorize Hollywood labour at the micro-level of practices and the macro-level of structure. An intersectional feminist approach to production studies examines how the incorporation of AI into
Hollywood film and television production practices will most strongly impact creative workers who are already marginalized in the industry