Daniel Fuentes, award for the best Master's Thesis in the High-Performance Computing Master's program

Daniel Fuentes, researcher at CiTIUS (a center co-financed by the European Union through the Galicia Feder Program 2021–2027), has just received the award for the best Master's Thesis (TFM) of the Master's in High Performance Computing, taught by the universities of A Coruña and Santiago de Compostela in collaboration with the Galicia Supercomputing Center (CESGA).

The work, titled Efficient multi-stage alignment of multispectral images using a multi-GPU algorithm, presents a new method that allows for the highly accurate alignment of multispectral orthomosaics (images created from multiple aerial photographs captured by drones with different color and light sensors), taken on different dates over Galicia's river ecosystems. The explored technique allows images obtained at different times to be superimposed with maximum detail, facilitating the comparison of specific areas and detecting subtle changes in the landscape, such as vegetation evolution, soil erosion, or the impact of infrastructure near rivers. This process is essential for assessing the health of natural ecosystems, studying the effects of climate change, or planning more precise conservation actions.

The method combines two complementary algorithms and takes advantage of parallel computing in supercomputers with multiple processors and graphics cards, such as CESGA's Finisterrae, which accelerates the processing more than twenty times compared to the sequential version.

The method combines two complementary algorithms, each specialized in a part of the image alignment process, and leverages the potential of parallel computing in supercomputers with multiple processors and graphics cards, like CESGA's Finisterrae. This type of infrastructure allows tasks to be distributed across hundreds of computing cores working simultaneously, multiplying processing speed and significantly reducing the time required to analyze large volumes of data. With this approach, the system manages to accelerate processing more than twenty times compared to the traditional sequential version, maintaining high precision in the results and opening up new possibilities for the automated study of territory from aerial images.

The work was supervised by Dora Blanco and Álvaro Ordóñez, both researchers at CiTIUS, and represents a notable example of how high-performance computing can contribute to environmental research and the monitoring of natural ecosystems.