GPU-accelerated 3D visualisation and analysis of migratory behaviour of long lived birds

Bird, Daniel (2021) GPU-accelerated 3D visualisation and analysis of migratory behaviour of long lived birds. Doctoral thesis, University of East Anglia.

[thumbnail of 2022BirdDPhD.pdf]
Download (31MB) | Preview


With the amount of data we collect increasing, due to the efficacy of tagging technology improving, the methods we previously applied have begun to take longer and longer to process. As we move forward, it is important that the methods we develop also evolve with the data we collect. Maritime visualisation has already begun to leverage the power of parallel processing to accelerate visualisation. However, some of these techniques require the use of distributed computing, that while useful for datasets that contain billions of points, is harder to implement due to hardware requirements. Here we show that movement ecology can also significantly benefit from the use of parallel processing, while using GPGPU acceleration to enable the use of a single workstation. With only minor adjustments, algorithms can be implemented in parallel, enabling for computation to be completed in real time.

We show this by first implementing a GPGPU accelerated visualisation of global environmental datasets. Through the use of OpenGL and CUDA, it is possible to visualise a dataset containing over 25 million datapoints per timestamp and swap between timestamps in 5ms, allowing for environmental context to be considered when visualising trajectories in real time. These can then be used alongside different GPU accelerated visualisation methods, such as aggregate flow diagrams, to explore large datasets in real time. We also continue to apply GPGPU acceleration to the analysis of migratory data through the use of parallel primitives. With these parallel primitives we show that GPGPU acceleration can allow researchers to accelerate their workflow without the need to completely understand the complexities of GPU programming, allowing for orders of magnitude faster computation times when compared to sequential CPU methods.

Item Type: Thesis (Doctoral)
Faculty \ School: Faculty of Science > School of Computing Sciences
Depositing User: Chris White
Date Deposited: 08 Aug 2022 12:32
Last Modified: 08 Aug 2022 12:32


Downloads per month over past year

Actions (login required)

View Item View Item