eScholar OntarioTech University
Not a member yet
1901 research outputs found
Sort by
Hyperspectral image compression using implicit neural representations
No full textHyperspectral images (HSI) capture the full electromagnetic spectrum for each pixel in a scene. They often hold hundreds of channels per pixel, providing significantly more information compared to a comparably sized RGB color image. As the cost of obtaining these images decreases, there is a need to create effective ways for storing, transferring, and interpreting hyperspectral data. In this thesis, we develop several neural compression-based methods for hyperspectral images. Our methodology relies on transforming hyperspectral images into implicit neural representations (INR), specifically neural functions that establish a correspondence between coordinates and features. We use a multilayer perception (MLP) network with sinusoidal activation functions that “learns” to map pixel locations to pixel spectrum for a given hyperspectral image. This representation thus acts as a compressed encoding of this image, and the original image is reconstructed by evaluating this network at each pixel location. In the other variation of using implicit neural representation to compress hyperspectral images, a sampling scheme is introduced to achieve better compression times while keeping decoding errors low. In our other method, instead of explicitly saving the weights of the implicit neural representation, the modulations that are applied to a base network that has been meta-learned are recorded. These modulations serve as a compressed coding for the hyperspectral image. An assessment of the proposed approach was conducted using four benchmarks: Indian Pines, Jasper Ridge, Pavia University, and Cuprite. The proposed method is evaluated against sixteen other schemes ((1) JPEG, (2) JPEG2000, (3) PCA-DCT, (4) PCA-JPEG2000, (5) MPEG, (6) X264, (7) X265, (8) PCA-X264, (9) PCA-X265, (10) FPCA-JPEG2000, (11) 3D-DCT, (12) 3D-DWT-SVR, (13) WSRC, (14) HEVC, (15) RPM, and (16) 3D-SPECK.) for hyperspectral image compression, and according to the Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) metrics, the method developed in this study achieves state-of-the-art compression rates at low-bit rates. We also used a large hyperspectral image dataset, compressed it using our methods, and compared our results with JPEG and MPEG. Finally, we conducted task-aware hyperspectral image compression, in which regions are chosen according to a task, and hyperspectral images are compressed using our proposed method
The impact of vaping/cigarette smoking on physical activity among youth in Canada: a longitudinal analysis of COMPASS data from 2021-2023
Full text linkBackground: This study examined the association between exclusive vaping/smoking status and meeting the Canadian Moderate to Vigorous Physical Activity (MVPA) guidelines (60 minutes/day) and the average weekly minutes of MVPA, and how these associations differed by gender.
Methods: A longitudinal analysis of a sample of high school students from British Columbia, Alberta, Ontario, and Quebec who participated in the Canadian COMPASS study from 2021-2023.
Results: Students who exclusively vape (OR = 1.16, 95% CI [1.01, 1.33]) and those who smoke (OR = 1.28, 95% CI [1.03, 1.58]) were more likely to meet the MVPA guidelines at follow-up. Exclusive vaping was positively associated with average weekly MVPA minutes (β = 46.30, 95% CI [12.10, 80.50]). The results differed among gender subgroups.
Discussion: Differences between unadjusted and adjusted models suggest that confounding and Sampson's paradox influence observed associations. Further longitudinal research is needed to explore long-term trends between vaping/smoking and physical activity
Language, power, and representation: developing a framework for digital best practices in autism discourse
Full text linkThe complex and evolving autism narrative is shaped by diverse actors, with digital platforms and autism organizations playing a critical role in advocacy and community engagement. Yet, autistic perspectives often remain underrepresented due to structural imbalances and top-down communication approaches, underscoring the need to assess three principles: inclusivity, accessibility, and credibility of online discourse. In this study, the websites of 14 Canadian autism advocacy organizations were critically analyzed to evaluate their alignment with these principles. Informed by Habermas’s Communicative Action Theory, a qualitative approach employing critical discourse analysis was used to examine website content and design. The organizations were selected through a multi-step process that used internet traffic ranking tools to ensure representation across provinces and territories. Each website was assessed for inclusivity, accessibility, and alignment with advocacy goals, drawing on key metrics such as the presence of autistic self-advocates, accessibility features, and transparent communication practices. The findings revealed considerable variability in website quality, with noticeable gaps in accessibility and the meaningful inclusion of autistic voices. To address these gaps, the CLEAR Framework (Clarity, Logic, Evidence, Accessibility, Representation) was developed and applied as an evidence-based tool rooted in universal design and neurodiversity principles. By emphasizing straightforward language, coherent messaging, credible evidence, accessible formats, and genuine autistic representation, the framework operationalized theoretical principles into actionable criteria, offering a structured tool for evaluating and refining how organizations communicate their missions, values, and practices. Ultimately, this research provides insights for guiding autism discourse, policy, and practice while also laying the groundwork for further investigations into equitable and inclusive digital advocacy across diverse contexts
Multimodal trajectory prediction-enhanced motion planning for autonomous driving in unsignalized intersections
Full text linkUnsignalized urban intersections pose significant challenges for autonomous vehicles due to ambiguous right-of-way rules and need to anticipate diverse human driving behaviors. This thesis addresses the problem of safe and efficient motion planning under such uncertainty by coupling a learning-based multi-modal trajectory prediction model with an optimization-based planner. Specifically, we propose graph neural network (GNN) architecture that captures spatial and relational interactions among road users, enabling prediction of multiple plausible future trajectories for each agent. These predictions inform a model predictive control (MPC) planner that enforces the most likely trajectories as hard constraints while treating alternate behaviors as soft constraints. This hybrid approach allows the ego-vehicle to reason over behavioral uncertainty in a principled manner. We evaluate framework on real-world intersection scenarios from the INTERACTION dataset, including four-way stops and unprotected left turns. Results demonstrate improved safety and comfort over single-trajectory baselines, with zero collisions and significantly smoother control profiles
An autoethnography of double consciousness and educational exclusion: Black womanhood, cultural erasure, and the search for belonging
Full text linkThis autoethnography explores my experience as a Black woman navigating the Canadian education system from childhood to my current educator role. Grounded in critical race theory and the concept of double consciousness, the study examines how race, gender and class interest shape identity, belonging and academic possibility. Through personal narrative and scholarly engagement, I reflect on the internalized silences, shifting identities, and systemic inequalities that shaped my sense of self across educational spaces.
While the educational trajectories of both my mother and daughter provide necessary context, this autoethnography centers my journey, illustrating how structural barriers, cultural erasure, and misrecognition defined my education and early academic career. My daughter’s experiences within today’s education system, including a health crisis nearly a decade ago, even more effectively illustrate how institutional practices continue to neglect the nuances of Black student life. Moreover, these experiences highlight an urgency for intersectional and humanizing approaches to student support.
Through the interconnectedness of memory, critical self-awareness, and theory, this paper challenges the myth of meritocracy and calls for a more relational, humanizing, and responsible understanding of educational equity
Exploring the lived experiences of racialized parents accessing pediatric rehabilitation services: descriptive case studies
Full text linkThis multiple descriptive case study utilized semi-structured interviewing to describe the intersection of race and disability among racialized parents (n=3) who have children with disabilities in a pediatric rehabilitation context. Through the within case analysis the results of this study revealed that racialized parents who have children with disabilities experience various barriers when trying to access services. The cross-case analysis revealed four main themes (a) ableism and formal diagnoses, (b) program wait times and scarcity, (c) the need for diverse service providers, and (d) desire for cultural support groups. The results from this study demonstrate that the experiences of racialized parents are complex, that their broader life experiences affect the way they approach pediatric rehabilitation, and that they may be experiencing a “double burden” effect
Investigation of novel electroactive morphing concepts for aerodynamic performance increase of an A320 wing through High-Fidelity numerical simulation
No full textThe present thesis aims at investigating bio-inspired concepts of morphing wings for greener aviation through aerodynamic performance increase. The work was conducted under the European project BEALIVE “Bioinspired Electroactive multiscale Aeronautical LIVE skin”. This research was carried out in collaboration between Ontario Tech University and IMFT. High-Fidelity numerical simulations were conducted in the Navier Stokes Multi-Block solver around the Airbus A320 wing in the subsonic regime for Reynolds number one million. Novel electroactive morphing concepts were explored through slight deformation and vibration of the trailing edge region for the design of a “live-skin”. Bio-inspired by fish scales and bird feathers, this concept consists of an innovative moving interface between the lifting structure and the surrounding turbulence by means of large DoFs of the actuators composing the morphing system. The Organised Eddy Simulation turbulence modelling was used to capture the turbulent coherent structures interacting with the chaotic turbulence. A large parametric study was performed with regard to a constant and time-modulation (“wobulation”) of the vibration frequency, enabling the detection of optimal morphing parameters. The actuation amplitude variation was investigated in the form of a sinusoidal spanwise travelling wave (STW) for different frequencies and wavelengths. Advanced spectral and wavelet analysis unveiled the presence of natural frequencies of the Kelvin-Helmholtz vortices, developed along the shear layers, as well as of the von Kármán vortex structures. Emphasis has been given to the evaluation of the spanwise undulations and vortex dislocations of the predominant vortex rows associated with secondary instability. Optimal vibrations with a constant amplitude and frequency were found to reduce drag up to −4%, increase lift up to +3%, and increase lift-to-drag ratio up to +5%. STW was found to provide a similar performance with simultaneous reduction of aerodynamic forces fluctuations up to −50% and reduction of noise sources associated with predominant modes up to 14 dB
Empowering young minds: exploring the impact of STEAM education through the Engineering Design Process (EDP) on female elementary students
Full text linkEarly exposure to STEAM experiences plays a valuable role in shaping elementary learners’ foundational engagement, STEM perceptions, and developing global competency skills. However, inconsistent learning opportunities highlight the need for inclusive pedagogical approaches that foster active engagement and motivation, particularly among girls. This systematic review examines the influence of the Engineering Design Process (EDP) on STEAM activities, identifying factors that affect learner engagement, the role of creativity in art-based tasks, and inclusive learning experiences for elementary-aged girls. Findings indicate that elementary girls engage more actively in relevant tasks that offer opportunities for creative expression, storytelling, or addressing issues of equity and social justice. This review, based on qualitative research, concludes that EDP-infused STEAM activities enhance student agency, artistic expression, and motivation by incorporating real-world relevance, providing multiple entry points to STEM content, and creating a supportive learning environment. Overall, while EDP-integrated STEAM activities show promise in supporting elementary girls’ interest in STEM, further research is needed to assess their long-term impact on STEM engagement and career pathways
Reforming EPZ framework for new nuclear reactors (SMR): an open-source and climate-aware approach
Full text linkThe research develops a weather-sensitive method for determining Emergency Planning Zones which applies to Small Modular Reactors and Pressurized Water Reactors while adapting to geographical conditions. The analysis examined severe accident such as Long- term Station Blackout (LSTBO), focused on how climate change increases risks related to things like flooding and severe snowfall. This study evaluates fixed-radius EPZ models because it conducts radiological dispersion simulations with RASCAL v4.3.4 to determine Total Effective Dose Equivalent (TEDE) and Thyroid Committed Dose Equivalent (CDE) outcomes across different seasons (summer, winter, monsoon) as well as analyzing year- long Meteoblue meteorological data. Python code was used to extract, analyze and visualize how dispersion changes each month following seasonal effect for EPZ evaluation. SMR-type nuclear reactors consistently deliver offsite doses that are lower than LPWRs according to research findings which confirm safety is maintained with smaller Emergency Planning Zone boundaries. SMR EPZ boundaries with radii smaller than 2 to 5 km become possible when using EPA and CNSC TEDE distribution thresholds. The findings indicate that small modular reactors (SMR) that adopt reformed emergency planning areas show greater flexibility and acceptance, thus can be implemented both in urban environments and remote locations with comparatively less societal upheaval to communities residing nearby. The research leads to the creation of a publicly accessible updated spatial framework for EPZs which improves both regulatory adaptability and public trust in new nuclear power systems
Development and evaluation of wind tunnel testing methodology for ADAS camera perception in rain
Full text linkAdvanced Driver Assistance System (ADAS) technologies are rapidly improving to enhance road safety and reduce accidents. However, adverse weather, particularly rain, continues to degrade sensor perception and effectiveness. Despite this, few studies address sensor degradation due to rain, with no established standards for benchmarking sensor performance loss. The objective of this thesis is to develop a methodology that surpasses conventional spray-based approaches in realism, allowing for controlled, repeatable, and quantifiable evaluation of sensor performance in rain. This thesis develops VeRSA, the most realistic indoor rain simulation system in open literature, now adopted commercially. Using VeRSA, camera image quality and object detection under dynamic rain are benchmarked, revealing key limitations in existing metrics. These findings enable the creation of rain-degraded datasets to enhance detection by retraining neural networks. Finally, a novel mathematical model is derived and validated to correlate rainfall with image degradation, establishing a foundation for predicting perception degradation