A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Identification of Urban Green Spaces in Lisbon using Semantic Segmentation
Dissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Data Science ; Urban green spaces are a crucial component of sustainable urban planning, providing numerous benefits to the environment and residents. The United Nations' Sustainable Development Goals (SDGs) underscore the importance of making public urban green spaces accessible, which makes it essential to effectively identify and manage these areas. In this study, the use of deep learning methods, particularly convolutional neural networks (CNNs), was explored to identify urban green spaces in Lisbon city through semantic segmentation. Two CNN architectures, U-Net and FPN, both equipped with a ResNet-50 encoder, were evaluated to determine their effectiveness in detecting urban green spaces from satellite images. The input data included open-source resources such as Sentinel images and ground truth data from an open data portal. The results reveal that the top-performing model achieves an average test Intersection over Union (IoU) of 0.347 and an average test F1-score of 0.51492, indicating moderate performance. This is largely because the training set lacks images with high cloud cover, while the test set includes them. Therefore, the visual results show that the model performs well on images with minimal cloud cover, suggesting it is reliable for identifying urban green spaces under clear conditions. The model's limitations in handling high cloud percentages highlight room for improvement, but its potential to contribute to sustainable urban planning is evident.
Identification of Urban Green Spaces in Lisbon using Semantic Segmentation
Dissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Data Science ; Urban green spaces are a crucial component of sustainable urban planning, providing numerous benefits to the environment and residents. The United Nations' Sustainable Development Goals (SDGs) underscore the importance of making public urban green spaces accessible, which makes it essential to effectively identify and manage these areas. In this study, the use of deep learning methods, particularly convolutional neural networks (CNNs), was explored to identify urban green spaces in Lisbon city through semantic segmentation. Two CNN architectures, U-Net and FPN, both equipped with a ResNet-50 encoder, were evaluated to determine their effectiveness in detecting urban green spaces from satellite images. The input data included open-source resources such as Sentinel images and ground truth data from an open data portal. The results reveal that the top-performing model achieves an average test Intersection over Union (IoU) of 0.347 and an average test F1-score of 0.51492, indicating moderate performance. This is largely because the training set lacks images with high cloud cover, while the test set includes them. Therefore, the visual results show that the model performs well on images with minimal cloud cover, suggesting it is reliable for identifying urban green spaces under clear conditions. The model's limitations in handling high cloud percentages highlight room for improvement, but its potential to contribute to sustainable urban planning is evident.
Identification of Urban Green Spaces in Lisbon using Semantic Segmentation
2024-11-07
203776526
Theses
Electronic Resource
English
Urban Green Spaces , Satellite Imagery , Semantic Segmentation , Convolutional Neural Networks , Deep Learning , SDG 3 - Good health and well-being , SDG 11 - Sustainable cities and communities , SDG 13 - Climate action , Domínio/Área Científica::Ciências Naturais::Ciências da Computação e da Informação
DDC:
710
Towards healthier cities: Urban Green Spaces (UGS) in the neighbourhood Benfica, Lisbon
| TIBKAT | 2021
Evaluating the Cooling Potential of Urban Green Spaces to Tackle Urban Climate Change in Lisbon
| DOAJ | 2019