Publicação
Automatic driving: 2D detection and tracking using artificial intelligence techniques
| Resumo: | Road accidents are estimated to be the cause of millions of deaths and tens of millions of injuries every year. For this reason, any measure that reduces accidents' probability or severity will save lives. Speeding, driving under the influence of psychotropic substances and distraction are leading causes of road accidents. Causes that can be classified as human since they all come from driver errors. Autonomous driving is a potential solution to this problem as it can reduce road accidents by removing human error from the task of driving. This dissertation aims to study Artificial Intelligence techniques and Edge Computing networks to explore solutions for autonomous driving. To this end, Artificial Intelligence models for detecting and tracking objects based on Machine Learning and Computer Vision, and Edge Computing networks for vehicles were explored. The YOLOv5 model was studied for object detection, in which different training parameters and data pre-processing techniques were applied. For object tracking, the StrongSORT model was chosen, for which its performance was evaluated for different combinations of its components. Finally, the Simu5G simulation tool was studied in order to simulate an edge computing network, and the viability of this type of network to aid autonomous driving was analysed. |
|---|---|
| Autores principais: | Pinto, José Miguel Fernandes Madeira |
| Assunto: | Autonomous driving Artificial intelligence Machine learning Computer vision Edge computing Condução autónoma Inteligência artificial Aprendizagem de máquina Visão por computador Computação de borda |
| Ano: | 2022 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade do Minho |
| Idioma: | inglês |
| Origem: | RepositóriUM - Universidade do Minho |
| Resumo: | Road accidents are estimated to be the cause of millions of deaths and tens of millions of injuries every year. For this reason, any measure that reduces accidents' probability or severity will save lives. Speeding, driving under the influence of psychotropic substances and distraction are leading causes of road accidents. Causes that can be classified as human since they all come from driver errors. Autonomous driving is a potential solution to this problem as it can reduce road accidents by removing human error from the task of driving. This dissertation aims to study Artificial Intelligence techniques and Edge Computing networks to explore solutions for autonomous driving. To this end, Artificial Intelligence models for detecting and tracking objects based on Machine Learning and Computer Vision, and Edge Computing networks for vehicles were explored. The YOLOv5 model was studied for object detection, in which different training parameters and data pre-processing techniques were applied. For object tracking, the StrongSORT model was chosen, for which its performance was evaluated for different combinations of its components. Finally, the Simu5G simulation tool was studied in order to simulate an edge computing network, and the viability of this type of network to aid autonomous driving was analysed. |
|---|