Digital twins of cities require accurate, up-to-date 3D building models, yet producing them still depends heavily on manual work. This project asks the intern to help automate part of our urban digital twin pipeline using deep learning and multi-modal geospatial data.
Working within ASIL, the intern will focus on reconstructing detailed (LoD2) 3D building models from drone and aerial imagery, LiDAR, and optical satellite data. Depending on interests and progress, the work may emphasize automated roof-structure and roofline extraction, fusion and registration of 2D images with 3D point clouds, or detecting changes between data captured at different times. The models feed directly into active lab projects, including a high-accuracy 3D digital twin of the Port of Saint John and broader smart-city and climate-adaptation applications such as flood mapping and urban heat-island analysis.
Over the 12 weeks the intern will review relevant literature, prepare and label geospatial datasets, implement and train deep-learning models (mainly in Python with PyTorch or TensorFlow), run and evaluate experiments, and visualize results in a GIS or 3D environment. They will work alongside graduate students and meet regularly with me to set goals and discuss progress.
By the end of the internship the intern will have hands-on experience across the full geospatial-AI workflow, from raw imagery to a validated 3D model, and a concrete contribution to a real digital-twin deployment. Strong outcomes can lead to a co-authored publication and a pathway toward graduate studies at UNB through the Globalink Graduate Fellowship.
Research area, student roles & skills
Research area: My research sits at the intersection of artificial intelligence, urban digital twins, and multi-modal geospatial mapping. In the Advanced Spatial Information Lab (ASIL), we develop deep-learning methods that turn drone, aerial, and satellite imagery, together with LiDAR point clouds, into accurate 3D city models and living digital twins of the built environment. Core themes include automated LoD2/LoD3 building reconstruction, multi-sensor data fusion and registration, and urban change detection. A central goal is using these digital twins to support climate-change adaptation, including infrastructure damage assessment, flood mapping, and urban heat-island analysis.
Student roles: The intern will be an active member of the Advanced Spatial Information Lab and contribute directly to our urban digital twin research rather than observing from the sidelines. Specific responsibilities include: (1) reviewing recent literature on 3D building reconstruction, multi-sensor fusion, and change detection to understand the state of the art; (2) collecting, organizing, cleaning, and labelling geospatial datasets drawn from drone flights, aerial and satellite imagery, and LiDAR; (3) implementing, training, and tuning deep-learning models in Python; (4) designing and running controlled experiments, then quantitatively evaluating accuracy against ground-truth data; (5) visualizing and interpreting results in GIS or 3D-modelling software; and (6) documenting methods and findings in clear written reports and lab presentations. The intern will meet with me regularly to set objectives and review progress, and will collaborate closely with graduate students and other lab members, including occasional participation in fieldwork or drone data-collection campaigns where schedules allow. We will tailor the precise focus to the intern's strengths and interests within the broader project. The role is designed to give a senior undergraduate a complete, authentic research experience: framing a problem, building and testing a solution, and communicating results. Reliability, good communication, and willingness to ask questions matter more than prior expertise, and I provide close mentorship throughout the 12 weeks. Promising interns will be encouraged to develop their contribution toward a conference paper or journal article and to consider returning for graduate study at UNB.
Skills required: Applicants should have a background in geomatics, remote sensing, computer science, software/computer engineering, electrical engineering, or a related field. Required: solid programming skills in Python and comfort working with data. Strongly preferred: exposure to deep learning (PyTorch or TensorFlow), image processing or computer vision, and familiarity with GIS, photogrammetry, or LiDAR/point-cloud data. Experience with Linux and version control (Git) is an asset. Most important are curiosity, attention to detail, and the motivation to learn quickly and work independently within a collaborative research team. No prior digital-twin experience is required.
2. Automatic detection of fluviomarine dunes from bathymetric data
We developed a deep learning method based on Mask-RCNN to detect fluviomarine dunes in a bathymetric model. Given the limited training data, we developed an approach to augment the dataset and perform transfer learning. Another challenge is ensuring the model's generalization.
The input data was obtained using image analysis methods. One limitation of the approach is that dune contours are sometimes poorly defined. The project involves implementing an approach to better define dune contours. This would improve the quality of the data provided to the training network and allow for more robust metrics.
The network was trained on data from the St. Lawrence River. The network generalizes well to different areas of the river, but it has not yet been tested in other marine areas. Its limitations will need to be studied in other coastal or marine environments.
Research area, student roles & skills
Research area: In geomorphology and environmental sciences, identifying landforms from images or terrain models helps us understand morphological processes or analyze the impact of climate. However, landforms are often vague shapes that lack clear boundaries. Boundaries are therefore difficult to characterize. Deep learning methods are therefore effective methods for this problem. However, they require large amounts of labeled data to train networks. We are therefore investigating methods, based on simulation or data augmentation, that limit this data dependency.
Student roles: The student will first work on improving the input datasets. The dunes are delimited by a footslope. These footslopes are represented in the images by polygonal lines. Some footslopes are poorly positioned. To improve the quality of the dataset, the student will implement an active contour model to better adhere to terrain characteristics. Active contouring is a method for detecting contours in images based on energy minimization. They will identify an image energy criterion, based on gradient or curvature, characterizing the foot slopes. They will use these new contours to produce a new training dataset and train the instance segmentation network. The results will be compared to assess the effectiveness of the approach. Finally, depending on the time remaining, the student will also test their approach on data from another area to assess the robustness of the segmentation method.
Skills required: The student must have training in a field related to geomatics or data science. They must have programming skills. Prior knowledge of image processing and prior experience in machine learning are advantages.
3. Conception d’une application de découverte du campus
L’équipe de recherche est multidisciplinaire, avec des membres experts en géomatique, en santé et en expérience utilisateur, et cherche à développer un outil pour aider au bien-être des personnes étudiantes, recommandant des espaces de confort sur le campus. Une application, DéambUL, permettant de découvrir le campus de l’université Laval en proposant des points d’intérêt (œuvres d’art, services) et des parcours autour de différents thèmes, comme la nature ou la santé, a déjà été développée. Le système repose sur un modèle permettant l’ajout simple de points d’intérêt et de parcours afin de mettre à jour l’application rapidement. Pour l’instant, les informations sont spatiales et ne sont pas temporelles alors que l’application pourrait servir à promouvoir des événements ou proposer des parcours en fonction des saisons. Des fonctionnalités pourraient aussi être ajoutées afin de faciliter la recherche de points d’intérêt en fonction du temps et de la position. Une première partie du travail consistera à étendre le modèle conceptuel pour intégrer la temporalité. Les points d’intérêt pourront être actifs à certains moments (ponctuellement ou périodiquement) et leur description pourra aussi varier dans le temps. Par exemple, les boisés du campus peuvent servir pour des promenades à pied ou en raquettes, ou les jardins peuvent avoir des attributs visuels ou olfactifs variant avec les saisons. Le nouveau modèle conceptuel sera implanté avec de nouvelles fonctionnalités de recherche associées à la temporalité et aux variables sensorielles des lieux.
Research area, student roles & skills
Research area: Je m’intéresse à la modélisation spatiale et à l’analyse spatiale. Cela inclut les algorithmes et les structures de données mais aussi la conception de bases de données spatiales et des systèmes d’information géographique. Un problème est la conception de modèles de données et de systèmes répondant aux exigences des utilisateurs. Il s’agit alors de définir des applications traitant la donnée spatiale efficacement tout en offrant des outils pertinents aux utilisateurs.
Student roles: La personne étudiante devra proposer un nouveau modèle conceptuel intégrant la temporalité. Elle proposera ensuite un modèle logique qu’elle implantera pour valider la conception. Elle proposera aussi de nouvelles fonctionnalités permettant de recommander des points d’intérêt ou des parcours en fonction de la position des utilisateurs et du temps. AInsi, elle interviendra sur toutes les étapes du projet, de la conception à la validation.
Skills required: La personne étudiante devra avoir une formation dans un domaine comme la géomatique ou la géographie intégrant des compétences en systèmes d’information géographique. Le développement de l’application se fera dans l’environnement ArcGIS, DéambUL étant publié comme un service ArcGIS on line. Une expérience en conception de bases de données spatiales est un avantage.
4. Detection of submarine dunes with unsupervised learning
We aim to develop a method for identifying dunes based on the detection of their critical points. These critical points are local maxima, saddle points, and local minima. Point clustering methods will be tested to group critical points belonging to the same dune. To achieve this, graph analysis methods will be used. The graph will be constructed by considering the nearest neighbours of each point: a critical point is connected to its closest neighbours. Then, three approaches will be tested: a geometric approach, HDBSCAN, a spectral approach based on the spectral analysis of the Laplacian matrix of the graph, and a community detection approach using the Louvain method. The performance evaluation of these approaches will be carried out by assessing their results with respect to the parameters specific to each method and by comparing them with an available ground truth. Memory costs and computation time will also be taken into account. All the tests will be conducted on a bathymetric model of the St-Lawrence river.
Research area, student roles & skills
Research area: In geomorphology and environmental sciences, identifying landforms from images or terrain models helps us understand morphological processes or analyze the impact of climate. However, landforms are often vague shapes that lack clear boundaries. Boundaries are therefore difficult to characterize. Deep learning methods are therefore effective methods for this problem. However, they require large amounts of labeled data to train networks. We are therefore investigating methods based on unsupervised learning that could limit this dependency to labelled data.
Student roles: The student will be required to design and develop the code needed to cluster critical points for each method. They will test the performance of the methods on the provided datasets. Finally, they will write a report discussing the performance of each method.
Skills required: The student should have strong skills in data analysis and possibly in geometric modeling. Algorithms already available in libraries such as Scikit-learn or NetworkX will be used. Programming experience is strongly recommended, as the student will need to write, run, and potentially optimise their scripts.
5. Error measurement in a digital elevation model modelled by an hexagonal grid
The research project aims to evaluate the quality of digital elevation models (DEMs) generated using a hexagonal grid and to compare them with DEMs produced using a regular grid. The DEMs are constructed from a lidar point cloud. Two case studies are considered: a point cloud acquired over the Université Laval campus, where the terrain is relatively flat, and a point cloud acquired in the Montmorency Forest, where the terrain is undulating and covered by boreal forest. Ground control points are available to assess elevation error.
The DEM will be generated by interpolating the elevation at the center of each cell from the lidar points. Three interpolation techniques will be evaluated: binning, linear interpolation based on triangulation of the point cloud, and inverse distance weighting. This evaluation will also consider parameter variations, such as the type of aggregation or the number of neighbors. The comparison will also be performed with respect to cell size. Finally, recommendations will be made regarding the relevance of using hexagonal grids for terrain modeling.
Research area, student roles & skills
Research area: High-resolution digital elevation models (DEM) are mainly produced from airborne lidar point clouds, in the form of regular grids. First, grids provide a uniform representation, whereas lidar point density varies greatly. Second, they are easier to handle. Some studies have shown that hexagonal grids are an interesting alternative for slope computation or river network mapping. However, there are relatively few studies comparing the quality of regular grids and hexagonal grids.
Student roles: The student will be responsible for implementing the processing workflows used to generate the DEMs. The application must allow the workflows to be executed while varying parameter settings. The student will also be required to assess elevation errors using a set of control points. Finally, at the end of the project, the student will prepare a report presenting their recommendations.
Skills required: A background in spatial analysis or numerical analysis is essential for the project. Programming experience is strongly recommended. The student is expected to be able to implement processing workflows using existing Python libraries.
6. Hidden Figures: Mapping Flood Exposure Inequities Across River Basins in Southern Ontario
Floods are among the most destructive natural disasters, with far-reaching consequences for communities worldwide. The increasing frequency and intensity of flood events, driven by climate change, underscore the urgent need for improved flood prediction, preparedness, and response strategies. However, many flood management efforts rely on traditional modelling approaches that emphasize rainfall-runoff relationships and binary floodplain delineations. While these methods offer valuable insights, they often lack the spatial and social granularity needed to capture real-time flood dynamics and the unequal burden of flood risk across communities. This project seeks to bridge this gap by integrating advanced satellite technology, machine learning, and community engagement to develop a comprehensive flood risk assessment framework for the Black Creek River Basin. By combining data-driven hazard modelling with measures of social vulnerability, we aim to generate high-resolution maps of flood vulnerability, hazard, and risk that enhance disaster preparedness and reveal disparities in flood exposure, particularly among racialized and marginalized communities. The methodology will leverage satellite-derived observations to characterize flood extent and dynamics, while machine learning algorithms will be trained to identify patterns and predict flood-prone areas with improved spatial precision. These technical outputs will be combined with socioeconomic and demographic indicators to construct a layered understanding of risk that accounts for both physical exposure and community resilience. Community engagement will ground the analysis in local knowledge and ensure that findings are relevant and actionable for affected populations. Recognizing these inequities is critical to ensuring that flood mitigation strategies are scientifically robust and socially just. By centring both technological innovation and social equity, this project aims to develop tools and insights that support more effective, inclusive, and responsive flood management, ultimately strengthening vulnerable communities' capacity to anticipate and withstand future flood events.
Research area, student roles & skills
Research area: I am a physical geographer interested in rivers, their processes, and corridors across scales. My overarching research focus is understanding the dynamics and drivers of river processes and their implications on river corridors in a rapidly changing world. As against the concept of looking at rivers as a mere conveyor of water, I consider rivers as corridors, that is, spaces of interaction between atmosphere, (a)biota and their floodplain and beyond. To do this, I combine field-based methods, modelling, remote sensing, and artificial intelligence to generate insights that connect fine-scale processes with large-scale patterns.
Student roles: The student will play a central role in developing and implementing the integrated flood risk assessment framework for the Black Creek River Basin. Working within an interdisciplinary research team, the student will take ownership of key technical and analytical components while contributing to the project's broader goals of advancing scientific understanding and social equity in flood management. A primary responsibility will be to acquire and process satellite imagery to map flood extent and characterize flood dynamics. The student will work with optical and radar (SAR) data, applying remote sensing techniques to extract meaningful information about flood hazards across the basin. This will involve using platforms such as Google Earth Engine and writing scripts in Python or R to automate data processing workflows. The student will also develop and train machine learning models to identify flood-prone areas and predict flood risk at high spatial resolution. This includes preparing training datasets, selecting appropriate algorithms, validating model performance, and refining outputs to ensure accuracy and reliability. In parallel, the student will integrate socioeconomic and demographic data to assess social vulnerability, combining physical hazard information with measures of community exposure and resilience. This work will culminate in high-resolution maps of flood vulnerability, hazard, and risk that highlight disparities in flood exposure among marginalized communities. The student will contribute to community engagement activities, helping to incorporate local knowledge into the analysis and to communicate findings to stakeholders in accessible formats. They will also document their methods and results, contributing to reports, presentations, and potential academic publications. Throughout the project, the student will collaborate closely with academic supervisors and partner organizations, participating in regular meetings, sharing progress, and incorporating feedback. This role offers valuable hands-on experience at the intersection of geospatial science, machine learning, and environmental justice.
Skills required: The ideal candidate will have a background in geography, environmental science, civil or environmental engineering, hydrology, geomatics, data science, or a related discipline. The student should have a strong foundation in geospatial analysis and be comfortable working with Geographic Information Systems (GIS) software such as ArcGIS or QGIS. Experience with remote sensing and satellite imagery analysis is highly desirable, including familiarity with platforms such as Google Earth Engine and with processing optical or SAR (radar) data for flood mapping. The candidate should have programming proficiency, particularly in Python or R, for data processing, automation, and statistical analysis.
7. Intégration des capteurs imageurs sur un ASV à faible coût
The project proposes the continued development of an autonomous hydrographic drone (ASV) platform dedicated to acquiring spatial data from various sensors, within the framework of Phase 5 of the REPER Ocean program. This platform was designed prioritizing the use of low-cost materials and sensors to offer an accessible, flexible solution adapted to user needs. It represents an innovative approach to acquiring hydrospatial data in various contexts, including coastal, lacustrine, and port environments, as well as areas that are difficult to access, shallow, or where data is scarce.
The work carried out to date has addressed several challenges related to sensor integration, system stability, and the platform's autonomous operation during acquisition missions. The various tests performed have demonstrated the robustness of the developed system and its ability to conduct hydrographic surveys autonomously.
This new implementation phase primarily focuses on integrating a high-performance, low-cost GNSS antenna to improve the accuracy, reliability, and quality of the ASV's positioning. This improvement will enable better georeferencing of the collected data and increase the accuracy of the generated hydrographic products, which are essential for mapping, port infrastructure management, planning interventions on berths, and studying natural environments.
A significant component of the project will also involve field validation of the GNSS integration during real-world data acquisition campaigns. These trials will allow for the evaluation of the system's performance, positioning stability, and the impact of this technology on the overall quality of surveys conducted by the ASV in various aquatic environments. Finally, ongoing documentation of the developments will be maintained to promote open access to the knowledge produced.
Research area, student roles & skills
Research area: My area of research concerns the acquisition and processing of geospatial data, particularly in geomatics engineering and surveying. My projects are divided into three axes, the first being the acquisition and preparation of acquired data, the second the formalization of knowledge enabling the automation of information extraction, and the third an analysis and modeling of the quality of acquired data. This project is mainly focused on the first axis, namely the integration of sensors on a low-cost data acquisition platform.
Student roles: At the start of the project, the student will receive training on previous developments on the ASV as part of its continuous improvement process. They will then need to familiarize themselves with the existing mechanical, electronic, and software components by performing comprehensive system functionality tests. One of the main tasks will be integrating the high-performance GNSS system into the platform, taking into account constraints related to weight, power supply, connectivity, and data synchronization. The student will be required to propose a suitable technical solution, implement it, and then conduct a critical evaluation of the integration. Field tests will be carried out to validate image quality, transmission stability, and compatibility with other onboard sensors. The goal is to have an operational system capable of capturing imagery data to complement hydrospatial data. In parallel, the student will be responsible for documenting each step of the integration process, updating technical guides, and writing detailed methodological reports for the scientific and technical community. These meticulously written documents will be published open access. The intern will also have the opportunity to participate in field campaigns to acquire bathymetric data and images in lake or coastal environments, thus contributing to the generation of cartographic products. They will be integrated into the REPER 3D laboratory research team, which fosters a collaborative, interdisciplinary approach to spatial representation. Active participation in team meetings is expected, as well as the ability to suggest improvements, receive constructive feedback, and adjust their work accordingly. The project strongly values autonomy, initiative, and creativity in a multidisciplinary environment at the interface of geomatics, robotics, and environmental observation.
Skills required: The student must have basic knowledge of programming and sensor integration. The student must be autonomous and be able to quickly appropriate the documents generated in the previous phase of the project, as well as the platform already developed. He/she must have the writing skills to formalize sensor integration and testing. Experience in embedded systems integration and programming using raspeberry pi is an asset for completing the project. The student must be open to participating in data acquisition and testing missions with the platform outside the company.
8. Intégration tridimensionnelle des capteurs hydrospatiaux à partir des nuages de points LiDAR
This research project is a continuation of a previous project aimed at optimizing the geometric integration of on-board sensors on a mobile hydrospatial acquisition system. It relies on point clouds from static terrestrial LiDAR sensors as the basis for estimating, with adequate uncertainty for geospatial applications, the relative position of sensors (e.g. GNSS antennas, cameras, LiDAR, MBES, inertial units) installed on a mobile platform. This approach is intended as an alternative to conventional total station georeferencing methods, which are often time-consuming, costly and dependent on a well-cleared working environment. The main aim of the project is to further develop a method for geometrically calibrating sensors based on high-density LiDAR data acquired in outdoor, uncontrolled environments. The approach involves extracting geometric features of the platform and sensors from the point cloud, in order to reconstruct the exact spatial configuration of the sensors by geometric adjustment. This method aims to ensure accuracy equivalent to, or even better than, that achieved by traditional approaches, while simplifying field constraints and reducing the need for external equipment. The project involves 3D data processing, geometric analysis and assessment of the quality of the adjustment obtained.
Research area, student roles & skills
Research area: My area of research concerns the acquisition, processing and use of geospatial data, particularly in surveying and geomatics engineering. My projects are divided into three axes, the first being the acquisition and preparation of acquired data, the second the formalization of knowledge enabling the automation of information extraction, and the third an analysis and modeling of the quality of acquired data. This project lies at the intersection of the first and third axes, since it concerns data acquisition systems and their three-dimensional modeling for the extraction of calibration parameters.
Student roles: The student will be trained in the principles of geometric integration of on-board sensors, mobile acquisition systems and the use of LiDAR point clouds for spatial calibration. He/she will be required to appropriate existing data sets from previous campaigns, including high-density point clouds and associated acquisition parameters. His role will be to extract geometric primitives such as planes, corners or edges from these data, and use them as references to estimate the position and relative orientation of sensors on the mobile platform. The student will conduct a series of geometric fitting tests to assess the feasibility and accuracy of this method, comparing it to reference data obtained by total station or other traditional approaches. The various processing, calibration and evaluation stages will be rigorously documented, with clear, reproducible protocols. Particular attention will be paid to the robustness of the proposed methods in the face of different acquisition environments and sensor configurations. The student will take part in research team meetings, present progress regularly, propose methodological adjustments and adapt to feedback received. This project will enable students to strengthen their skills in 3D data processing, sensor geometry and error modeling. It will give him hands-on experience of the challenges involved in integrating embedded data into mobile mapping systems, while exploring alternative approaches to conventional calibration methods. Autonomy, methodological rigor and scientific curiosity will be essential to the success of this applied research project.
Skills required: The student must have knowledge of geometry and trigonometry, geomatics, 3D modeling and spatial data manipulation. They must be autonomous and able to quickly assimilate the datasets provided, as well as the software tools used to use and model data from point clouds. Experience with imagery and LiDAR or 3D modeling data formats and programming is an asset. The student must also be willing to participate in data processing and validation activities in the field or controled environment.
9. Monitoring and assessing blue-green infrastructure using remote sensing
Blue-green infrastructure, as a nature-based solution, is increasingly implemented in urban areas to improve stormwater management and provide ecosystem services such as biodiversity enhancement and urban heat island mitigation. Traditional on-site monitoring and assessment methods are important for understanding long-term performance and maintenance needs, but they are often labour-intensive and costly. This highlights the need for efficient monitoring and analytical approaches for blue-green infrastructure systems.
This research project aims to develop and use remote sensing-based approaches to efficiently monitor blue-green infrastructure at large scales and understand potential factors that influence the performance of blue-green infrastructure systems. As part of a larger research initiative, this internship will focus on collecting, digitizing, processing, and analyzing a subset of geospatial and design data from Toronto.
During the internship, the student will synthesize information, process raw data, and analyze geospatial datasets. The student will gain hands-on experience in geospatial analysis, remote sensing, GIS, programming, and statistical analysis, while strengthening their knowledge of blue-green infrastructure design, environmental monitoring, and data analysis.
The outcomes of this research are expected to improve the monitoring, design, and planning of blue-green infrastructure. The project will also contribute to broader efforts to enhance urban ecosystem services, resilience, and sustainability.
Research area, student roles & skills
Research area: Dr. Liao’s research focuses on green infrastructure, water treatment, water resource management, plant-soil-water interactions, environmental monitoring and remediation, and remote sensing. Her work aims to optimize the monitoring and design of nature-based systems to improve water quality, plant performance, and ecosystem functions. She also investigates the design and application of sustainable, engineered materials for the removal of both traditional and emerging contaminants from water and the environment, while protecting ecosystem health. In addition, Liao employs remote sensing and data science approaches (e.g., meta-analysis) to understand, assess, and optimize vegetation, water, and environmental performance and management at multiple spatial scales.
Student roles: The student will contribute to a broader research project on remote sensing-based monitoring of blue-green infrastructure in urban environments. The role will involve collecting, organizing, digitizing, processing, and analyzing geospatial and design data, as well as supporting data synthesis, spatial analysis, and interpretation of findings. The student will help identify factors influencing blue-green infrastructure performance and contribute to the development of more efficient large-scale monitoring methods. This role will provide hands-on experience in geospatial analysis, remote sensing, GIS, programming, and environmental monitoring research.
Skills required: - Academic background in geography, computer science, environmental science, environmental/civil engineering, geomatics, GIS, remote sensing, or a closely related field. - Strong interest in blue-green infrastructure and geospatial or remote sensing applications in environmental research. - Previous experience with GIS software, remote sensing analysis, spatial data processing, programming (e.g., Python), or statistical analysis would be an asset. - Strong analytical, organizational, communication, and teamwork skills.
In the realm of mass-market location-based services, smartphones have emerged as the predominant terminals due to their widespread usage, portability, and cost-effectiveness. The majority of research in smartphone navigation has concentrated on GPS, particularly following the availability of raw GPS measurements from Android smartphones in 2016. Urban navigation presents distinct challenges arising from issues such as imprecise GPS positioning, dynamic traffic conditions, and complex intersections. Instances of smartphone navigation solutions placing users on the wrong side of the street are not uncommon, let alone achieving lane-level accuracy in urban settings, which is unattainable solely through GPS. Typically, visual observations from smartphone cameras mounted on vehicle dashboards can be employed in visual odometer or visual navigation based on high-definition maps, with deep learning playing a crucial role. With the capability to handle non-linear observations, factor graph optimization has been applied in various applications, including sensor calibration and sensor fusion. This project outlines a comprehensive research initiative focused on enhancing smartphone navigation in urban areas through factor graph based integration of multiple onboard sensors, including cameras and GPS, coupled with the utilization of deep learning and Android APIs. The intended integration aims to mitigate navigation errors, furnish users with precise
Research area, student roles & skills
Research area: My research interests include sensor fusion, localization and mapping, estimation theory, machine learning for navigation, navigation in challenging environment, intelligent traffic system, robotics, and computer vision in pose estimation. He is aiming to explore the power of sensor fusion and machine learning in localization and mapping to improve the accuracy, safety, and reliability of autonomous driving.
Student roles: Students will work on the following three modules, 1) GPS positioning module with Android smartphone. 2) Deep learning based end-to-end visual navigation module. 3) Factor graph optimization based sensor fusion module.
Skills required: The applicants are expected to have a strong background in Navigation, Geomatics, Remote Sensing, GIS, Computer Science, or Robotics. Excellent programming skills, preferably in Java or C++/C or Python, are desired.
11. SpatioPhon : Plateforme géoinformatique dédiée aux paysages sonores
Supervisor: Frédéric Hubert
University: Université Laval (Québec campus)
Location: Québec, Québec
Start date: 2027-05-02 (flexible)
Disciplines: Geomatics, Geography, Engg-Computer, Land Information, Engg-Software
La conservation des paysages en milieu urbain constitue un enjeu majeur pour nos sociétés actuelles, dans un contexte où la pression urbaine influence notre relation à l’environnement, notamment en matière de bien-être et de qualité de vie. Afin de promouvoir et de préserver la signature patrimoniale unique des paysages, il devient primordial d’intégrer spatialement les dimensions sonores et acoustiques dans la cartographie d’un territoire. Ce projet vise ainsi à valoriser le patrimoine acoustique du campus de l’Université Laval, tout en identifiant les zones de quiétude. Plus précisément, le principal objectif est de concevoir et de développer une infrastructure géoinformatique permettant de caractériser le paysage de sonore du campus en mobilisant des solutions géoinformatiques et acoustiques. Plusieurs phases devront être réalisées :
1. Inventaire des propriétés acoustiques et des sources sonores du campus. Des appareils acoustiques (enregistreur audio, sonomètre) pourront être utilisés pour réaliser des collectes de données sur le terrain. L’analyse des données audio devra permettre d’identifier les types de sons (transport, biodiversité, vent, etc.) en utilisant des méthodes d’intelligence artificielle (Audiotimm, YAMNet, PANNs).
2. Collecte de données géospatiales pour caractériser le territoire selon ses propriétés acoustiques, et produire de nouveaux jeux de données avec le GeoAI, tels que l’identification d’obstacles (ex. murets, talus) à la propagation des sons.
3. Production d’une cartographie du bruit à l’aide d’outils open source (NoiseModeling, OpeNoise).
4. Conception d’une base de données spatio-phoniques permettant de stocker les résultats d’analyse des données audio géoréférencées, les cartes de bruit, mais également toute information contribuant à la caractérisation du paysage sonore du territoire.
5. Développement d’une application web de géovisualisation de cartes spatio-phoniques, connectée à la base de données. Plusieurs représentations cartographiques du paysage sonore devront être proposées.
6. Identification et représentation des zones de quiétudes et/ou des trames blanches à l’échelle du campus.
Research area, student roles & skills
Research area: Ce projet s'inscrit à la croisée de plusieurs domaines : acoustique, géomatique et informatique. L'acoustique vise à étudier les propriétés acoustiques des territoires, l’identification des phonies et les méthodes propagation des sons dans l'environnement. La géomatique vient en appui à l'acoustique pour fournir et traiter des données géospatiales contextuelles destinées aux inventaires et analyses, ainsi que pour produire des cartes de qualité adaptées. L'informatique joue le rôle d’interface en permettant la mise en place des infrastructures et des outils de calculs appropriés.
Student roles: Dans le cadre de ce stage, l’étudiant sera amené à: • Acquérir des connaissances relatives aux données spatiales, aux paysages sonores, aux bases de données géospatiales et à la géovisualisation; • Étudier des solutions open source d’intelligence artificielle dédiées à la classification des données audio (ex. Audiotimm, YAMNet); • Collecter des données sur le terrain et procéder à leur analyse (types de sons) • Collecter les données géospatiales et appliquer des techniques de GeoAI pour identifier et délimiter les éléments manquants du territoire (obstacles) • Produire une carte de bruit du campus à l’aide d’outils tels que NoiseModelling ou OpeNoise. • Concevoir une base de données géospatiales pour stocker les résultats d’analyse des données audio tout en conservant leur localisation. Ces résultats devront être complétés par les propriétés acoustiques du territoire et les constituants du territoire (route, bâtiments, obstacles, etc.). • Concevoir et développer une application de cartographie web du paysage sonore, basée sur les données issues de la base de données. • Définir et produire différentes cartographies l’environnement sonore, tout en délimitant les zones de quiétudes et/ou trames blanches. • Produire un rapport final décrivant de manière explicite les démarches réalisées dans le cadre de ce travail.
Skills required: • Connaissance minimale en géomatique (Système d'information géographique/SIG, géoinformatique) est considérée comme un plus. • Connaissance de certains langages de programmation (Java, Python, Javascript, HTML, CSS). • Connaissance des technologies GitHub et Docker est considérée comme un plus. • L'expérience dans les réalisations de stages techniques et/ou de projets en lien avec le projet sera considérée avec attention. • Une compétence en lien avec des outils d’apprentissage automatique / IA sera considérée comme un plus. • Un attrait et un intérêt pour le sujet devront être clairement démontrés.
12. Vers la conception d’un jumeau numérique 3D de l’environnement sonore
Supervisor: Frédéric Hubert
University: Université Laval (Québec campus)
Location: Québec, Québec
Start date: 2027-05-02 (flexible)
Disciplines: Geomatics, Geography, Engg-Computer, Land Information, Engg-Software
The acoustic environment is primarily studied from the perspective of noise, to the detriment of the soundscape. Noise is a major issue in today’s global society, as it is one of the most harmful factors affecting public health (sleep disturbance, stress, well-being, heart attacks). Today, the province of Quebec has the political will to study, understand and combat it more effectively. Noise mapping is the primary tool for studying and understanding how sounds propagate, as well as for carrying out simulations (such as the addition of noise barriers). Beyond the extensive research in this field and open-source initiatives, certain limitations exist, such as the production of 2D maps of environments, whereas the 3D component offers significant potential for improvement in terms of simulation, visualisation and understanding of how pollution spreads in urban environments for non-experts.
The aim of this project is to design and develop a workflow for producing a 3D digital twin of the acoustic environment. This will enable the creation of a geoinformatics infrastructure for modelling, disseminating and visualising the acoustic environment (environmental noise and soundscapes) using free open-source solutions. At the heart of this solution, we will need to include a module for the automatic generation of 3D environmental noise modelling, using existing propagation models, road traffic simulation solutions and acoustic sensors. Furthermore, the temporal component will need to be studied to enable its use within the application. The components of the soundscape must also be integrated into the solution (i.e. perception, sound sources, phonies). Finally, various acoustic, geomatics and IT solutions must be explored and potentially utilised to implement this infrastructure, such as Cesium.js, QGIS, Unreal, NoiseModeling, GitHub and Docker.
Research area, student roles & skills
Research area: This project is at the intersection of several fields: acoustics, geomatics and computer science. The acoustics component focuses on the acoustic properties of the terrain and the ways in which sound propagates through the environment. Geomatics supports the acoustics by providing and processing geospatial data to support propagation calculations, as well as by producing high-quality, adapted maps using, for example, interpolation methods. Computer science acts as a buffer by establishing the appropriate infrastructure and computational tools.
Student roles: • Learn the fundamentals of spatial data, noise mapping and 3D geovisualisation; • Examine the open-source NoiseModelling solution designed for modelling and mapping environmental noise; • Examine an open-source solution designed for road traffic simulation (e.g. MATSim, SUMO) ; • Identify techniques and solutions for 3D dissemination and visualisation on the web (tiling and virtual globes); • Identify existing 3D sound environment mapping applications (web-based or otherwise), highlighting the different methods for modelling and representing these phenomena in 2D and 3D; • Evaluate and test, where appropriate, 3D mapping solutions with a view to visualising the sound environment in 3D; • Design and develop a 3D noise modelling module to feed a virtual globe such as Cesium.js; • Implement the complete infrastructure for modelling, dissemination and visualisation; • Define and implement different methods for 3D mapping of the sound environment (points on facades, by floor, by building, on the ground, etc.); • Investigate temporal noise modelling for real-time dissemination using simulation tools such as MATSim or SUMO and data from acoustic sensors; • Produce a final report describing in detail the steps carried out during this project.
Skills required: • Basic knowledge of geomatics (Geographic Information Systems/GIS, geoinformatics) is an advantage. • Knowledge of certain programming languages (Java, Python, JavaScript, HTML, CSS). • Knowledge of GitHub and Docker technologies is an advantage. • Experience in technical internships and/or projects related to the project will be given careful consideration. • Skills in noise mapping would be an advantage. • A clear interest in and enthusiasm for the subject must be demonstrated.
13. Évaluation empirique de l’impact de la qualité des données géospatiales dans le cadastre 3D multisource
Ce projet de recherche vise à répondre à un enjeu stratégique émergent dans le domaine de la gestion du territoire : l’évaluation de l’influence de la qualité des données géospatiales sur la fiabilité des modèles de cadastre 3D multisource. Alors que les villes deviennent de plus en plus complexes et que les infrastructures occupent simultanément des espaces en surface et en sous-sol, le cadastre tridimensionnel s’impose comme une solution innovante pour représenter avec précision les propriétés, les bâtiments, les réseaux souterrains et les droits associés aux volumes occupés.
Cependant, la création de modèles cadastraux 3D fiables repose sur l’intégration de données provenant de sources variées, dont la qualité géométrique et sémantique peut être très variable. Cette diversité représente un défi majeur pour assurer la précision, la cohérence et la valeur opérationnelle des modèles générés. Le projet a donc pour objectif de mesurer concrètement l’impact des caractéristiques des données géospatiales, telles que la résolution, la précision, la complétude et l’incertitude, sur la qualité finale des représentations cadastrales 3D.
Dans le cadre de cette recherche appliquée, la personne étudiante contribuera à l’acquisition, la préparation et l’analyse de données issues de différentes technologies de pointe, notamment la photogrammétrie aérienne et terrestre, les levés LiDAR, ainsi que des données géospatiales ouvertes telles que les modèles numériques de terrain. Différentes stratégies d’intégration et de combinaison de sources seront testées afin d’identifier les approches offrant le meilleur compromis entre précision, coût et disponibilité des données.
Les résultats permettront de mieux comprendre les limites et les avantages des différentes sources géospatiales pour la production de cadastres 3D robustes. Cette recherche contribuera au développement de meilleures pratiques pour la modélisation territoriale, la gestion foncière intelligente et la planification urbaine.
Research area, student roles & skills
Research area: Mon domaine de recherche concerne l’acquisition, le traitement et l’utilisation des données géospatiales, en particulier en arpentage et en génie géomatique. Mes projets sont décomposés en trois axes, étant le premier l’acquisition et la préparation des données acquises, le deuxième la formalisation des connaissances permettant l’automatisation de l’extraction des informations et le troisième une analyse et la modélisation de la qualité des données acquises. Ce projet s’insère surtout dans le troisième axe, mais des éléments des deux premiers seront également fortement présents.
Student roles: Au début du projet, l’étudiant recevra une formation sur le cadastre 3D, la qualité des données géospatiales et les outils nécessaires à leur analyse. Il devra s’approprier des jeux de données multisources (LiDAR, photogrammétrie, relevés topographiques) afin d’évaluer leur qualité selon des critères comme la précision, la complétude ou la cohérence géométrique/topologique. À partir de cas d’étude, il mènera des tests comparatifs pour analyser l’impact de ces variations de qualité sur la fiabilité des représentations cadastrales 3D. Il devra documenter rigoureusement ses démarches et rédiger des rapports clairs, destinés à être partagés en libre accès. Intégré à une équipe de recherche multidisciplinaire, il participera activement aux réunions, présentera ses résultats et proposera des pistes d’amélioration. L’autonomie, la rigueur et l’esprit critique seront essentiels à la bonne réalisation de ce projet. Ce projet de recherche permettra à l’étudiant de consolider ses compétences en analyse spatiale, en modélisation 3D et en gestion de la qualité des données. Il offrira aussi une initiation concrète aux défis actuels liés à l’intégration de données hétérogènes dans des systèmes géospatiaux de nouvelle génération.
Skills required: L’étudiant doit avoir des connaissances en géomatique, en modélisation 3D et en manipulation de données spatiales. Il doit être autonome et capable de s’approprier rapidement les jeux de données fournis ainsi que les outils logiciels utilisés dans le cadre du projet de cadastre 3D. Il doit posséder des compétences en analyse spatiale et en évaluation de la qualité des données. Une expérience avec les formats de données d’imagerie et LiDAR, les logiciels SIG ou de modélisation 3D est un atout. L’étudiant doit aussi être disposé à participer à des activités de traitement de données et de validation sur le terrain.