Earth & Space Science

Permanent URI for this collection

https://hdl.handle.net/10315/27626

Browse

Now showing 1 - 20 of 185

Open Access
Resident Space Object Tracking for Space Situational Awareness
(2024-11-07) Kunalakantha, Perushan; Lee, Regina S. K.
This research presents several contributions aimed at improving the current state of Space Situational Awareness (SSA) using optical imagery. The first component involved the development of an image capture algorithm which was successfully used to acquire over 90000 optical images from the stratosphere, with hundreds of visually verified Resident Space Objects (RSOs). The second component involved the development of a novel RSO tracking algorithm which was able to detect 87% of the RSOs in an 878-image dataset at least once. The third component proposed an automated annotation framework and corresponding four-tool annotation suite to develop a 500-image dataset to train and test another RSO tracking algorithm. The final component involved the demonstration of a dual-purpose payload, performing RSO detection alongside an existing Attitude Determination (AD) algorithm which detected 11 unique RSOs in real-time during another stratospheric balloon mission.
Open Access
New Generation Stochastic Data-driven Calibration of the Accelerometers and Modelling of the Non-gravitational Accelerations in GRACE missions
(2024-11-07) Tzamali, Evangelia Myrto; Pagiatakis, Spiros
The Gravity Recovery and Climate Experiment (GRACE) and its successor, GRACE Follow-On (GRACE-FO), have significantly advanced our understanding of Earth's gravity field and its temporal changes by measuring subtle gravity variations caused by mass redistribution on and beneath Earth's surface. A key component of these missions is the use of geodetic-quality GPS receivers for precise orbit determination and 3D accelerometers, which measure non-gravitational forces like atmospheric drag and solar radiation pressure. However, challenges in using accelerometers must be addressed to ensure data and modeling accuracy. A primary challenge is the calibration of accelerometers, as inaccuracies can lead to significant errors in gravity field modeling and thermospheric density estimates of up to 10%, depending on the levels of solar activity. This study introduces a calibration method using matched filter techniques applied to accelerometer measurements and GPS-derived accelerations. The estimated calibration parameters show high stability with values close to 1, as expected from the ultra-sensitive electrostatic space accelerometers and proposed by other studies. Error assessment of GRACE and future mass change missions must consider accelerometer measurements as stochastic quantities with realistic error estimates. This involves identifying and quantifying measurement errors stemming from instrument noise, thermal effects, correlations and transient effects. This study generates a stochastic (weighted) 1B accelerometer dataset to investigate systematic errors from various sources like geomagnetic storms, temperature changes, and terminator crossings. Contrastingly to the official accuracy which is 〖0.1 nm/s〗^2 for the along-track and radial direction and 〖1 nm/s〗^2 in the cross-track, the proposed dataset being clean from spikes has an accuracy of 〖 10^(-3) nm/s〗^2. Accurate modeling of non-gravitational forces is essential for isolating gravitational signals and estimating drag, which is crucial for determining thermospheric densities. This study proposes a data-driven model for the dominant forces of solar and thermal radiation pressure and drag on the GRACE-C satellite, using the new 1B dataset produced in this study. A comprehensive analysis of residuals which are up to 〖2 nm/s〗^2 in the along-track, 〖0.5 nm/s〗^2 in the cross-track and up to 〖5 nm/s〗^2 in the radial direction, reveals disturbances with latitudinal dependency, especially during high geomagnetic activity.
Open Access
On the Impact of Land Use & Landcover Change on GHG Emissions using Advanced Remote Sensing Technology
(2024-11-07) Ituen, Ima-Obong; Hu, Baoxin
There is an interesting effect of the changing climate: it has been observed in Northern Ontario that the warming weather has lengthened the growing season. Consequently, there has been increased land use conversion from natural forests to agriculture in Northern Ontario to capitalise on the new economic opportunities resulting from longer growing seasons. However, the long-term effects of land conversion on soil carbon, nutrients, and greenhouse gas emissions (GHGs) are unknown. This research study leveraged advanced remote sensing technology to detect changes in land cover and land use. A major problem addressed in this work was the lack of landcover classification data which is current, extends over the entire study area, and is of high spatial and temporal resolution with which to analyse the study region properly. This study proposed a new change detection framework which uses the best available data for landcover classification and disturbance information, even in the event of scarce input or training data. Thus, an analyst can progressively adapt the available data sources to present a more accurate assessment of the disturbances. More accurate input data for the carbon models was shown in this study to decrease carbon and GHG estimates by between 15% and 32%. The study concludes with tools that an everyday user can utilize to examine their region. The applications can run for free on their devices, permitting users to discover the landcover changes in their region, and estimate forest loss or gain. The applications could serve as a bridge to discussions concerning the areas that are best suited to concentrate carbon sequestration efforts.
Open Access
Image Processing Techniques for Space Situational Awareness
(2024-07-22) Stewart, Michael Ian; Lee, Regina
With a renewed interest in space, it is paramount to understand the space environment; an act commonly referred to as Space Situational Awareness (SSA). Tracking and performing measurements on nearby orbiting objects; known as RSOs helps us better understand SSA. Image processing is a significant step in improving SSA capabilities, aiding in space object detection, tracking, and characterization. This thesis addresses gaps in RSO image processing, with a focus placed on developing specialized pipelines for the production of photometric measurements of Geosynchronous Equatorial Orbit (GEO) RSOs and the James Webb Space Telescope (JWST). Despite challenges like the unique Earth-Sun L2 Orbit of JWST and high noise issues, this study successfully produced calibrated photometric light curves of three targeted RSOs, advancing our understanding of photometric SSA methodologies.
Open Access
BIM based Energy Consumption Estimation using Data-driven model
(2024-07-18) Kiavarz Moghaddam, Hamid; Jadidi, Mojgan
Building Information Modeling (BIM) is undergoing rapid technological evolution in the building construction industry. Recently, employing BIM as a building 3D digital model in Building Energy Consumption Estimation (BECE) has gained momentum because of the enriched geometric and semantic information. Indeed, indoor BECE notably depends on the semantics, geometry (building elements and shapes), and topology information of the building's elements to recognize the spaces in a building with high energy demand. However, despite extensive studies on applying the BIM and Industry Foundation Classes (IFC) as an open standard data model for BIM in BECE analysis, employing the full potential of the BIM remains poor due to its data model complexity and incompatibility with BECE data-driven algorithms. There is a significant lack of building energy modeling in using the detailed geometry, semantic, and 3D topology information in BECE data-driven models. The objective of this dissertation is to develop an innovative and comprehensive framework called space-based precise building energy consumption estimation using BIM. In this research, a framework is developed to convert the IFC model into a space-based graph, including the geometry, semantic, and topology information on the proposed graph nodes and edges. The graph is compatible with the machine learning algorithm. A graph-based classification algorithm is suggested in this research to find critical spaces in the building for energy consumption. This research proposed a prescriptive model by integrating building energy simulation with optimization techniques, using BIM data and a Genetic Algorithm (GA) to develop a prescriptive model for indoor building design. The study focuses on space-based BECE analysis, leveraging BIM interoperability to recommend optimal solutions. The proposed model employs the value engineering method to balance energy consumption, functionality, and cost, providing engineers and designers with insights to optimize building performance effectively. This approach enhances energy efficiency and offers substantial design optimization solutions, bridging the gap between energy prediction and practical application in the architectural, engineering, and construction (AEC) industry. The outcomes of this study are conducive to contemporary data-driven models in BIM and indoor BECE analysis. This provides a comprehensive perspective on both present and prospective requirements for BIM in the estimation of building energy consumption. The study integrates various sectors, including architecture, construction, machine learning, ad 3D geospatial analysis, aiming to derive comprehensive and optimal solutions. Furthermore, it underscores the necessity for future multidisciplinary research by unfolding existing gaps and limitations.
Open Access
Neural Network Based Sliding Mode Control for Robotic Manipulator
(2024-07-18) Gomes Carmo, Ingredy Gabriela; Shan, Jinjun
Robotic manipulators are used in many applications. However, robotic arms are complex systems due to external disturbances, perturbations, and their coupled non-linear dynamics. This thesis aims to propose a robust control strategy for autonomous robotic manipulation. First, the trajectory tracking problem was introduced and an approach to overcome this issue using a sliding mode controller combined with a neural network is proposed. Then, the proposed approach is compared to classical and modern control methods including controllers from the literature to demonstrate the performance of the proposed controller. The proposed controller was then integrated with a grasp detection algorithm for an autonomous manipulation application. Simulations and hardware experiments were conducted to validate the performance of the proposed method.
Open Access
A Study on Depth Estimation and Digital Terrain Model Reconstruction for Mobile Mapping Systems
(2024-07-18) Naeini, Amin Alizadeh; Sohn, Gunho
Mobile Mapping Systems (MMSs) are advanced platforms for collecting precise geospatial data. They use technologies like LiDAR and digital imaging to gather detailed environmental information. MMSs have two main components: georeferencing and mapping. Georeferencing aligns digital data with coordinate systems using Global Navigation Satellite Systems (GNSS) for high accuracy. In areas where GNSS is unavailable, Simultaneous Localization and Mapping (SLAM) technology is used to map unknown environments and track MMS locations. The accuracy and density of depth maps are crucial for SLAM performance, affecting the system's ability to create maps and track positions. This dissertation addresses depth estimation challenges in MMSs by introducing the Double-stage Adaptive Refinement Scheme (DARS). DARS is designed to improve depth estimation in dynamic environments and can be integrated with pre-trained networks. It can also be extended to Panoramic DARS (PanoDARS) for SLAM applications using panoramic images, improving the HDPV-SLAM system's performance by addressing LiDAR depth sparsity and depth association issues. The thesis also explores reconstructing Digital Terrain Models (DTMs) from Digital Surface Models (DSMs), essential for accurate mapping within MMSs. DTMs provide elevation data for creating geospatial products like orthophotos, topographic maps, and 3D urban models. To achieve precise DTM reconstruction, a geospatially induced autoencoder called SB-SUBNET is proposed. This autoencoder uses two geospatial inductive biases that leverage the relationship DTM = DSM - nDSM. The first bias reconstructs the DTM by subtracting the network's output (nDSM) from its input (DSM). The second bias is a subtractive skip connection that integrates this geospatial information directly into the network, enhancing performance using inherent geospatial relationships in the data. While SB-SUBNET shows promising results, it alters all DSM values, including terrain values, affecting DTM accuracy. To address this, a new multi-task learning approach named DB-SUBNET was developed. This approach segments non-ground points and performs regression specifically for these points while preserving ground points. This method improves DTM reconstruction accuracy and ensures the resulting DTMs provide a reliable foundation for high-precision geospatial products. Enhanced DTM reconstruction is vital for creating detailed and accurate maps, supporting urban planning, infrastructure development, environmental monitoring, and disaster management.
Open Access
Intelligent Decision-Making for Autonomous Driving in Dynamic and Interactive Environments
(2024-07-18) Yuan, Mingfeng; Shan, Jinjun
The decision-making module is crucial for safe and efficient driving in autonomous vehicles (AVs). However, AVs face significant challenges in coexisting with human road users, making fast and optimal driving decisions, and operating in unknown traffic environments with only incomplete information. Unsignalized intersection and lane-changing scenarios are particularly representative of such challenges, which involves complex dynamic interactions. AVs need to assess and predict the driving preferences of nearby vehicles to optimize and adaptively adjust their own driving policies while considering uncertainties arising from incomplete observations. This dissertation investigates game-theoretic and learning-based methods to address these challenges. In this dissertation, a novel approach for integrating game-theoretic decision-making with deep reinforcement learning (DRL) is proposed to enable AVs to navigate unsignalized intersections using an onboard sensor. The game model predicts the surrounding vehicles' reactions to the ego-vehicle's movements without relying on coordination or vehicle-to-vehicle communication. The proposed algorithm employs cognitive hierarchy reasoning and a DRL algorithm to achieve a self-play training mode for getting a near-optimal driving policy in a realistic simulator before transferring to the real world. Second, this dissertation introduces a practical algorithm based on DRL for enhancing lane-changing decision-making, addressing low sample efficiency in DRL, and improving the generalization capability in Imitation Learning (IL). To narrow the gap between simulation and reality (sim-to-real gap), a digital twin platform is developed for simulating LiDAR sensing, model training, and algorithm evaluations. To tackle multi-objective optimization and imbalanced data concerns, a hierarchical decision-making framework is proposed, breaking down the complex decision-making problem into subtasks for improved convergence of driving policies. Third, a robust adaptive game-theoretic decision-making algorithm by utilizing receding horizon optimization and level-k game theory is presented. To reduce the potential safety risk arising from an inaccurate motion prediction of surrounding vehicles, the proposed approach can estimate driving aggressiveness of surrounding vehicles online. Then, the generated trustworthiness is used to formulate a safe, efficient, and robust adaptive driving policy. Additionally, a switching interaction graph is introduced into the adaptive level-k framework to reduce the computational complexity. Validation on both a high-fidelity simulator and hardware confirms the feasibility, effectiveness, and real-time performance of the proposed methods. Overall, this dissertation contributes novel approaches to address decision-making challenges in AVs. The integration of game theory and model-based/model-free optimization showcases the potential for improving safety and efficiency of AVs' operation in dynamic traffic environments.
Open Access
Smartphone precise positioning in urban environments using internal GNSS and IMU sensors
(2024-07-18) Yi, Ding; Bisnath, Sunil
The high level of signal noise and multipath effect errors cannot be neglected when it comes to smartphone-grade GNSS receivers and antennas, and along with frequent carrier-phase measurement discontinuities and losses, pose a challenge for advanced GNSS positioning techniques. In response to these challenges, this work is dedicated to addressing these smartphone positioning obstacles and aims to provide accurate, continuous, and resilient Positioning, Navigation, and Timing (PNT) solutions with a developed smartphone processing software. This research commences with the investigation of smartphone Precise Point Positioning (PPP) performance with external ionospheric constraints. To effectively utilize all satellite measurements in the absence of phase measurements, this work proposes a pseudorange-only measurement enhanced PPP method with single- and dual-frequency combinations. Moreover, to overcome the limitations of both PPP and Real-Time Kinematic (RTK) techniques, a novel PPP/RTK hybridization algorithm with smartphone Inertial Measurement Unit (IMU) integration is proposed. Additionally, this research explores the utilization of the available Galileo HAS corrections for smartphone navigation. Validated through a series of vehicle experiments conducted in realistic driving environments, the results demonstrate a substantial improvement in reducing horizontal rms, improving it from approximately 10 metres to less than 1.5 metres compared to traditional Single Point Positioning (SPP) solutions, showing a noteworthy improvement in smartphone-based mobile positioning for mass-market applications.
Open Access
Architectural Heritage Exploration and Visualization Using Interactive and Immersive Technologies
(2024-07-18) Albourae, Abdullah Taha; Armenakis, Costas
The Historic District of Jeddah (HDJ) of the Kingdom of Saudi Arabia, is facing many challenges to attain conservation and sustainability, due to natural disasters and the absence of appropriate digital recordings and documentation. Previous studies on world heritage conservation attempted to address these challenges while adopting the immersive and visual technologies separately or via merging one of them with another to conserve the heritage sites (e.g. BIM/GIS integration or generating GIS/BIM, integrating VR with BIM, or GIS). However, few of these studies handled the case of the historic heritage of KSA. This research aims to fill this gap by utilizing all these technologies together to create an immersive visualization system for heritage sites. The researcher proposes a novel method using immersive technologies (VR, AR, MR) and visualization methods (3D modeling) combined with Heritage Building Information Modelling (HBIM) and Heritage Geographic Information Systems (HGIS) for comprehensive conservation of cultural heritage sites in Saudi Arabia. The proposed method involves three phases: a) HBIM and HGIS generation: This involves creating 3D architectural models and spatial data using industry-standard collection tools, b) HBIM and HGIS Integration: The data from both phases is merged into a single georeferenced system for unified analysis, and c) Interactive User Experience: An interactive user environment is developed using a game engine, allowing users to explore the 3D model and access historical information. This research offers a multi-disciplinary approach to cultural heritage conservation, combining innovative informatics with technology for documentation, communication, and user engagement. The findings will enable users to virtually explore the heritage sites and track changes over time. This immersive experience fosters a deeper understanding and appreciation of the architectural heritage. The proposed method can be applied to other cultural heritage sites globally. Future studies could explore using 3D viewers for information retrieval and potentially using CityGML, a rich 3D geospatial data format, for even more comprehensive HBIM-HGIS integration. This would significantly contribute to the advancement of world heritage conservation efforts.
Open Access
Design, development and verification of a UAV-based GNSS payload for reflectometry
(2024-07-18) Talebi, Sogand; Bisnath, Sunil B.
This thesis focuses on the design, development, and verification of a Global Navigation Satellite System-Reflectometry (GNSS-R) payload tailored for Unmanned Aerial Vehicles (UAVs) utilizing Commercial Off-The-Shelf (COTS) components. UAVs serve as versatile aerial platforms capable of operating at diverse altitudes, thereby offering varying levels of spatial resolution suitable for localized remote sensing applications. A customized GNSSR payload was created and tested in a comprehensive flight campaign over diverse terrains. The collected data underwent post-processing to generate reflectometry products, such as Delayed Doppler Maps (DDM) and Delayed Waveforms (DWF). Comparative analyses were conducted on GNSS-R reflections from different surfaces to confirm the payload’s reflectometry capabilities. The research results offer valuable insights into the practical applications of GNSS-R technology in remote sensing, emphasizing its potential viability for various applications in this field.
Open Access
Image Classification and Initial Orbit Determination of Resident Space Objects (RSO)
(2024-07-18) Vallecillo Baires, Andrea Maria; Lee, Regina
The importance of Space Situational Awareness (SSA) research is steadily growing due to continuous launches of new technologies, leading to increased congestion and potential collisions in Earth’s orbit. Optical imagers, based in both ground and space, serve as important tools for observing, detecting, and studying resident space objects (RSOs). Optical imagers capture vast amounts of data that can be used for different applications in SSA research. However, manually inspecting and classifying these images for specific purposes is a time-consuming task. Implementing an automated image classification can streamline the labelling process for optical databases to expedite SSA research. Another significant aspect of SSA research and RSO tracking using optical images involves reliable object identification and Orbit Determination (OD). Angles-only Initial Orbit Determination (IOD) methods are often employed as a starting point to optimize the OD process. These advancements play a pivotal role in enhancing and contributing to SSA research.
Open Access
Integrative and Multi-scale Deep Learning for 3D Point Cloud Transmission Corridor Scene Segmentation: Noise Filtering, Attention-Fused Feature Integration, and Panoptic Network
(2024-03-16) Jameela, Maryam; Sohn, Gunho
LiDAR technology plays a crucial role in mapping natural and built environments across various civic and military applications. It enables the acquisition of high-density 3D point clouds with pulse repetition frequencies ranging from 100Hz to 2MHz. However, the increased overlap with atmospheric points has posed challenges in noise filtering and 3D point cloud quality. This dissertation proposes the Noise Seeking Attention Network (NSANet), a novel solution integrating psychological theories of feature integration and attention engagement. NSANet achieves a 4.10% increase in F1-Score and a 7.30% improvement in recall by employing multiscale context, global physical priors, and local spatial attention for noise filtering, surpassing previous techniques. The study explores the relationship between vegetation height and plantation guidelines, identifying spatial layout consistency in utility layouts and transmission objects. This insight drives the development of three semantic analysis approaches: Semantic Utility Network (SUNet), Fusion-Semantic Utility Network (Fusion-SUNet), and Panoptic-Semantic Utility Network (Pan-SUNet). Encouraged by the performance improvements of SUNet and Fusion-SUNet, Pan-SUNet achieves outstanding results, boasting a 94% F1-Score for pylons, 99% for ground, vegetation, and powerlines, and demonstrating high precision in 3D object detection. Experiments conducted on Teledyne Optech's Galaxy T1000 dataset, which features diverse voltage transmission lines, validate the effectiveness of Pan-SUNet, particularly when combined with the RandLA baseline. Significant improvements are observed, including an increase from 80% to 85% in F1-Score for pylons, 98% to 99% for ground, 93% to 97% for powerline, 75% to 78.3% for other objects, 86% to 88% for buildings, while maintaining 98.2% for high vegetation and 93% for medium vegetation. The key contribution of this research is a significant advancement beyond basic object classification, as it not only identifies the class of an object but also distinguishes between different instances of the same class. This instance segmentation is critical for utility network modelling and simulation. The research emphasizes the importance of external cues such as contextual reasoning, spatial cognition, and physical priors for multiscale fusion in scene understanding systems. Moreover, the study's adaptability to integrate proposed contributions into existing networks enhances their overall performance, making them network-agnostic.
Open Access
Integration of Airborne Laser Scanner and Optical Data Using Image Tie Points and the Quasi-Log-Polar Analytical Fourier Mellin Transform
(2024-03-16) Baumgaertner , Michael S. P.; Armenakis,Costas; Wang, Jianguo
LiDAR and camera have complementary data characteristics. Their combined usage allows for achieving better accuracy and enhanced inference about the environment than through the isolated use of the data of one sensor only. Co-registration is the process of spatial alignment of data from different types of sensors. It enables cross-aided sensor- and data calibration and the combined usage of the data. Automatic co-registration of airborne optical imagery and LiDAR data from large-area surveys has not been addressed, unless the datasets come with metadata or some prior knowledge about the scenery or have been pre-processed and converted into data with similar density and structure. In this dissertation, an innovative co-registration concept that uses image tie points, projected into object space, and Lidar points, is proposed and researched. Since 3D tie points are generated automatically with centimeter-level precision within the photogrammetric block adjustment, this approach has the potential of providing high accuracy and full automation. However, it must address that large area photogrammetric tie points provide extremely low point density compared to laser point clouds, and they describe different aspects of the scanned surface, since tie points coincide with radiometric corner locations on the visual surface whereas laser points locate anywhere on surface objects and on the ground. This research provides the following contributions: Firstly, the use of 3D photogrammetric tie points for co-registration of large-format sensor data is proposed and analyzed. Secondly, new methods for coarse- and fine registration in space domain are presented and evaluated. Thirdly, five methods for co-registration in frequency domain are researched. Lastly, the novel Quasi Log-Polar Fourier Mellin Transform (QAFMT) method is suggested as tool for automatic coarse co-registration of large area datasets. The learnings and potential use for other applications are discussed in the end. The performance of the QAFMT and its competing methods is analyzed using one simulated and two real-world datasets. While existing algorithms estimated 45 of 144 co-registration parameter pairs correctly (31% success), the QAFMT yielded a 7% improvement (50/144 accurate pairs, 38% success).
Open Access
Automated Plume Rise Measurement based on Deep Neural Networks using Video Images
(2024-03-16) Koushafar, Mohammad; Gordon, Mark; Sohn, Gunho
When a smokestack releases pollutants, the resulting plume cloud dissipates gradually and mixes with the surrounding air; it becomes neutrally buoyant and loses its vertical momentum. It is then carried downwind at a constant height called the Plume Rise or plume rise height. Plume rise affects how far pollutants are carried downwind, their deposition to the environment, and the amount of greenhouse gases mixed into the upper troposphere. Therefore, correctly calculating plume rise for the modelled dispersion of pollutants is of concern in air-quality transport models and local environment assessment cases. Recent studies have shown that the Briggs equations, which are a popular form of parameterization in models, significantly underpredict the plume rise. Modern computer vision methods allow the possibility of measuring plume rise under varied atmospheric conditions using video images. Most existing computer vision methods detect smoke clouds using an estimated bounding box without performing a segmentation down to the pixel level. Our proposed method can accurately detect and segment a plume cloud exiting from a smokestack and consider a hypothetical plume centerline based on an improved Deep Convolutional Neural Network (DCNN). We propose a Mask R-CNN model that can be applied for extracting the region of the plume cloud area of interest. Then, the proposed network is modified with our training dataset and used for detecting the hypothetical centerlines of the plume cloud. Finally, a comparative analysis was performed using meteorological data and smokestack measurements. This analysis involved comparing the plume rise and plume rise distance values estimated by the proposed framework with the obtained values from the Briggs parameterization equations. The outcomes confirmed a considerable underestimation of plume rise by the Briggs plume rise parameterization in the region. Using our model, the conventional plume rise model validation methods can be developed, and scientific grounds are provided for developing new physical models.
Open Access
Application of physics-informed neural network approach in soil moisture retrieval using GNSS reflectometry
(2024-03-16) Gavili Kilane, Narin; Bisnath, Sunil B.
This research aims to obtain soil moisture from reflected GNSS signals using physics-informed neural networks (PINN). GNSS reflectometry (GNSS-R) signals can be considered as a new remote sensing source to study soil moisture. Despite the high sensitivity between GNSS reflected signal power and soil moisture, the model between measurements and parameters is difficult to solve mathematically due to the complexity of the electromagnetic relationships. Although Neural Network (NN) algorithms have been applied successfully in GNSS-R soil moisture retrieval, neural networks are trained without respecting any laws of physics. In this work, a new framework referred to as “physics-informed neural networks (PINN)” was used which adds governing physical relationships between data parameters to neural networks to generate more robust models, with less data. The proposed research advances GNSS-R soil moisture estimations, exploiting Cyclone Global Navigation Satellite Systems (CYGNSS) satellite signals using PINN methodology. In PINN’s structure, reflected GPS signals from CYGNSS and land surface geophysical parameters are used as input features. Since reflected signal power variations are not only sensitive to changes in soil moisture, but also to changes in vegetation, surface roughness, soil texture, and elevation angle, the effects of land surface geophysical parameters involved in physical relationships are considered in the model. For reference data, soil moisture measurements of the International Soil Moisture Network (ISMN) were used in both training and validation. The proposed PINN model generates daily soil moisture values with a root mean squared error (RMSE) of 0.05 〖cm〗^3/ 〖cm〗^3, which is an improvement from 0.0774 〖cm〗^3/ 〖cm〗^3 for the underlying NN model due to adding physical models. Four different soil dielectric constant models (Dobson, Hallikainen, Mironov, and Wong models) have been used to investigate the impact of soil dielectric constant models as part of physical relations. The RMSE distinction and correlation coefficient difference of the best model (Hallikainen) and worst model (Mironov) is 0.02 and 0.13, respectively demonstrating PINN sensitivity to different soil dielectric constant models. Consequently, the soil dielectric constant model selection influences overall PINN results. Thus, calibration of soil dielectric models is necessary for GNSS-R soil moisture retrieval in the future.
Open Access
Small-Scale Gravity Waves generated by Hurricanes in the Upper Troposphere - Lower Stratosphere
(2024-03-16) Wang, Yuying; Pagiatakis, Spiros D.; Vergados, Panagiotis
This thesis presents a detailed study of the small-scale gravity waves in the upper troposphere-lower stratosphere region that are generated by hurricanes. The goal is to understand the spectral properties of the convectively generated gravity waves and their variations with respect to the hurricane intensity, using the high vertical resolution temperature measurements from the GPS-Radio Occultation (RO) and the fine horizontal resolution wind data from the ECMWF reanalysis dataset, with the application of the least squares spectral analysis and wavelet analysis. Our results show that the pure thermal forcing mechanism generates the primary waves at the surface level in an almost identical way, while the overshooting convection and the background wind filtering seem to weaken as the hurricane intensity drops. The obstacle effect appears to dominate in the lower stratosphere which further contributes to the wave field asymmetry. A two-layer convection model is considered to simulate the hurricane convection that innovatively connects the wave generation and dissipation mechanisms with the ambient background atmosphere.
Open Access
Image-Based Spatial Change Detection Using Deep Learning
(2023-12-08) Bousias- Alexakis, Evangelos; Armenakis, Costas
Image change detection is an invaluable tool in monitoring and understanding the built and physical environments and supporting decision-making. Many recent research approaches for automatic change detection have been based on Deep Learning (DL) techniques and especially on variations of Convolutional Neural Network (CNN) architectures. CNNs have achieved notable success thanks to their great representational capacity, straightforward training, and state-of-the-art performance in visual tasks. Nevertheless, CNNs, like most DL approaches, still face limitations relating to their reliance on extensive labelled datasets, the localization accuracy and detail of the predicted outputs, and the notable model performance degradation when the target data have different characteristics from the training data. This research contributes to the development and evaluation of novel DL methods and algorithms for automated image-based spatial change detection. It investigates novel architectures for enhanced model performance and ways to address the limited availability of labelled data and improve model generalizability. Two main approaches are investigated: (i) a direct approach, where both instances are fed into the DL model that is trained to output the prediction of the change map, and (ii) a post-classification approach, where change detection is performed based on each epoch’s semantic segmentation predictions. In both cases, novel, enhanced CNN architectures have been proposed that leverage the semantic information of objects’ boundaries to improve the accuracy of the model’s predictions. Furthermore, training frameworks inspired by self-ensembling and the Mean Teacher method were developed for semi-supervised learning and domain adaptation, attenuating the models’ reliance on large, labelled training datasets and improving their generalization performance. We evaluated the proposed approaches on multiple datasets using the precision, recall, F1 score, and Intersection over Union (IoU) metrics. The results indicate that both boundary-enhanced approaches lead to consistent, albeit marginal, benefits between 1 and 2 %. Notably, the proposed semi-supervised training framework for direct CD approximately matches the performance of the fully supervised approach while using only 20% of the available training labels. In domain adaptation, our post-classification approach significantly outperforms typical supervised methods, with the most notable gains in recall rate (>22%) and IoU (12.6%). These findings highlight the effectiveness of our techniques.
Open Access
An Examination of the ThermopiezoelectricEffect in Piezoelectric Actuators Via Finite Elements
(2023-12-08) Toledo, Rafael Chaves; Orszulik, Ryan
The application of piezoelectric actuators in smart structures is a rapidly developing field, especially in aerospace environments. Since thermal effects play an important role in aerospace environments, thermopiezoelectricity has been studied as it takes into account the thermal field in addition to the mechanical and electrical fields. As a result, the coupling effects among these three fields have to be considered, including the pyroelectric (change in the electric potential due to the presence of a thermal field) and electrocaloric (change in the temperature when an electric field is applied) effects. This thesis presents an examination of how these coupled effects can affect the performance of piezoelectric bender and stack actuators in varying external environments. More specifically, this thesis investigates the influence of the pyroelectric and electrocaloric effects on the positioning and dynamic performance of these actuators in static and dynamic cases by using a custom written finite element code that considers the three fully coupled field equations of thermopiezoelectricity.
Open Access
Visual-LiDAR Simultaneous Localization and Mapping
(2023-12-08) Ahmadi, SeyedMostafa; Sohn, Gunho
Simultaneous Localization And Mapping (SLAM) has garnered significant attention in robotics research over the years. While SLAM has demonstrated success, its application in mobile mapping systems (MMS) presents unique challenges. This study builds upon prior research (RPV-SLAM), extending its framework to enhance accuracy and perform boundary tests on a specific MMS, Maverick MMS. Our contribution introduces a novel SLAM approach termed HDPV-SLAM, addressing critical limitations encountered by the existing system. The first challenge addressed is the sparsity of LiDAR depth data, complicating its correlation with extracted visual features from RGB images. The second challenge stems from the lack of horizontal overlap between the panoramic camera and the tilted LiDAR sensor, causing difficulties in depth association. Furthermore, a comprehensive dataset named YUTO MMS is presented to the public. This dataset spans 18.95 km and was collected from diverse environments, including York University's Keel campus and Teledyne Optech headquarters building.