Research
Research
PhD Thesis
Supervisor
Manuel Isidro Martín-Ortíz, PhD.
Full-Time Principal Researcher C in the Faculty of Computer Science at the Benemerita Universidad Autónoma de Puebla (BUAP). Puebla, México.
Co-Supervisor
Full-Time Principal Researcher C in the Computer Science Department at the Instituto Nacional de Astrofísica, Óptica y Electrónica (INAOE). Puebla, México.
Honorary Senior Research Fellow in the Computer Science Department at the University of Bristol, in the UK.
Abstract
This thesis presents the development of continual learning strategies for visual localisation of a monocular camera onboard a drone in situations of GPS signal loss. The main objective was to design a backup localisation methodology capable of providing an approximate position while dynamically incorporating new information. To achieve this, three localisation schemes based on continual learning were proposed: topological localisation, hierarchical localisation, and progressive localisation, all using visual perception of the environment through aerial images. In addition, two continual learning strategies and a search mechanism were implemented to associate test images with their corresponding positions.
The experiments were conducted in controlled scenarios with real flight data, evaluating the performance of the proposed methods against discontinuous trajectories and diverse environments. The results demonstrated that the methodologies are able to recover approximate localisation at the image processing rate, confirming their usefulness as a backup system in the event of GPS failure. Unlike traditional approaches that require long training times and large datasets, this work explored the use of lightweight networks and binary architectures, capable of estimating the position in near real-time, at a rate close to the camera’s capture frequency.
The main contributions of this thesis include: (1) the development of continual learning strategies specialised for visual localisation; (2) the creation of a multi-model methodology that assigns positions along the flight trajectory; and (3) the use of lightweight models such as Support Vector Regression and Binary Networks, which can compete with deep architectures in terms of accuracy and efficiency. These advances provide a benchmark for the development of future visual backup localisation systems, particularly in environments where no external positioning infrastructure is available.
This research produced the following list of publications:
Peer-Reviewed Journal Papers Indexed in the Journal Citation Report:
A. A. Cabrera-Ponce, M. I. Martin-Ortiz, J. Martinez-Carranza. “Continual Learning for Topological Geo-localisation", Journal of Intelligent & Fuzzy Systems. Pre-press, pp. 1-13, April 5, 2023. https://doi.org/10.3233/JIFS-223627
Peer-Reviewed Journal Papers Indexed in the Emerging Sources Citation Index (ESCI):
A. A. Cabrera-Ponce, M. I. Martin-Ortiz, J. Martinez-Carranza. “Continual Learning via Multiple Support Vector Models for Localisation with a Single Aerial Image”. Unmanned Systems. 2025. https://doi.org/10.1142/S2301385026500172
Peer-Reviewed Journal Papers in other indexes:
A. A. Cabrera-Ponce, M. Martín-Ortiz, J. Martinez-Carranza. “Localización de una Cámara Monocular utilizando Métodos de Visión y Aprendizaje Profundo: Una Descripción General”, United Academic Journals (UA Journals), 2022. https://issuu.com/uajournals/docs/000015
Book Chapters and Peer-Reviewed International Conferences:
A. A. Cabrera-Ponce, M. I. Martin-Ortiz, and J. Martinez-Carranza. “A Review on Binary Networks for 6D Aerial Pose Estimation", Handbook of Intelligent Robots: Theory, Methods and Applications, Chapter. Taylor & Francis CRC Press, 2025. (Accepted)
A. A. Cabrera-Ponce, M. I. Martin-Ortiz, and J. Martinez-Carranza. “Continual Learning for Camera Localisation", Machine Learning for Complex and Unmanned Systems, Chapter. CRC Press, 2024. https://doi.org/10.1201/9781003385615
A. A. Cabrera-Ponce, L. O. Rojas-Pérez, M. I. Martin-Ortiz, and J. Martinez-Carranza. “Binary networks and continual learning for pose estimation from a single aerial image,” in 15th International Micro Air Vehicle Conference and Competition (IMAV2024). Bristol, United Kingdom, September, 2024. https://www.imavs.org /papers/2024/11.pdf
A. A. Cabrera-Ponce, M. Martín-Ortiz, J. Martinez-Carranza. “Hierarchical Continual Learning for Single Image Aerial Localisation", in 14th international micro air vehicle conference and competition (IMAV2023). Aachen, Germany, September 11-15, 2023. https://www.imavs.org/papers/2023/5.pdf
A. A. Cabrera-Ponce, M. Martín-Ortiz, J. Martinez-Carranza. “Multi-model continual learning for camera localisation from aerial images,” in 13th international micro air vehicle conference (IMAV2022), Delft, the Netherlands, 2022, p. 103–109. https://www.imavs.org/papers/2022/12.pdf
Other Peer-Reviewed Reports in International Events:
A. A. Cabrera-Ponce, M. Martín-Ortiz, J. Martinez-Carranza. “Aprendizaje profundo para localización aérea", Komputer Sapiens. Enero-Abril, 2025.
A. A. Cabrera-Ponce, M. Martín-Ortiz, J. Martinez-Carranza. “Discrete Hierarchical Continual Learning for Single View Geo-Localisation", Computer Vision and Pattern Recognition Conference: LatinX in AI (LXAI) Research Workshop (CVPR2023). Vancouver, Canada, 2023. https://doi.org/10.52591/lxai202306189
Master Thesis
Supervisor
Full-Time Principal Researcher B in the Computer Science Department at the Instituto Nacional de Astrofísica, Óptica y Electrónica (INAOE). Puebla, México.
Honorary Senior Research Fellow in the Computer Science Department at the University of Bristol, in the UK.
Abstract
The Global Position System (GPS) has become an essential sensor for applications and public systems, maritime systems, and aerial vehicles in the robotics area. Traditionally, autonomous flight in outdoor areas is possible thanks to GPS devices that enable the Unmanned Aerial Vehicle (UAV) to obtain its position in latitude and longitude coordinates. However, GPS may become unreliable when the drone flies in environments where the signal may get occluded. Malicious attacks may also compromise the GPS signal, aiming to block the signal or replace it with spurious data. Motivated by these scenarios, the proposed approach relies on a methodology to estimate the GPS position of a UAV using Convolutional Neural Networks (CNN) and a learning-based strategy. For the latter, we adopted two learning scenarios: 1) offline learning; 2) online learning, where we address the re-localisation problem and geo-localisation in a scenario where GPS devices fail.
We argue that our approach could be used as a backup system to return the UAV home when the GPS device is not working. We performed performance tests with aerial images and videos captured with the Drone Matrice 100 and Drone Parrot Bebop 2 in two scenarios with different trajectories to demonstrate our approach using a compact CNN and online training implementation. The experiments present suitable results; on average, CompactPN obtained an error of 2.60 to 6.16 metres with a speed of 107.88 fps. Likewise, our implementation based on the AR1* method obtained an accuracy of 0.7931 to 0.9282 with a performance speed of 127.33 fps for online training results.
This research produced the following list of publications:
Index Journals:
Aldrich A. Cabrera-Ponce, J. Martinez-Carranza. “Convolutional Neural Networks for Geo-Localisation with a Single Aerial Image”. Journal of Real-Time Image Processing. February, 2022. JCR Q3. https://doi.org/10.1007/s11554-022-01207-1
Peer-Reviewed International Conferences:
Aldrich A. Cabrera-Ponce, and J Martinez-Carranza. “Aerial Geo-Localisation for MAVs using PoseNet”. IEEE 5th Workshop on Research, Education and Development of Unmanned Aerial Systems (RED-UAS). IEEE, Cranfield, UK. November 2019.
Aldrich A. Cabrera-Ponce, Manuel Martin-Ortiz, and Jose Martinez-Carranza. "Continual Learning for Multi-camera Relocalisation". 20th Mexican International Conference on Artificial Intelligence (MICAI). Mexico City, Mexico. October 2021.
Bachelor Thesis
Supervisor
Full-Time Researcher B in the Computer Science Department at the Instituto Nacional de Astrofísica, Óptica y Electrónica (INAOE). Puebla, México.
Honorary Senior Research Fellow in the Computer Science Department at the University of Bristol, in the UK.
Abstract
Recently, unmanned aerial vehicles (UAVs) have grown in the research areas of aerial robotics, thanks to their onboard cameras. The videos captured with the cameras allow the exploration of environments for solutions such as object detection, surveillance and those requiring aerial photography for disaster inspection. However, such a task requires a correct joining of the images, in addition to real-time. The present thesis aims to perform the stitching of aerial images to generate a real-time mosaic, based on visual descriptor correspondences extracted from the images. The first stage consisted of the calculation of the ORB descriptors and their comparison based on the libraries of a SLAM system. The second stage consisted of joining these images by calculating their transformation matrix and their correspondences. These correspondences are important to show those points that coincide between a first image and the second one, combining them in a wide image in such a way that a mosaic is generated with the joined images. The system was tested in four outdoor environments with image capture at different heights and the results show that the system achieves the correct joining of the images to generate a mosaic in real-time.