Loading...
Fine-grained Permeable Surfaces Mapping through Parallel U-Net
Ogilvie, Nathaniel
Ogilvie, Nathaniel
Citations
Altmetric:
License
License
DOI
Abstract
Permeable surface mapping which mainly is the identification of surface materials that will percolate is essential for various environmental and civil engineering applications, such as urban planning, stormwater management, and groundwater modeling. Traditionally, this task involves labor-intensive manual classification, but deep learning offers an efficient alternative. Although several studies have tackled High Resolution Remote Sense image segmentation, the challenges in automated permeable surface mapping arid environments remain unexplored because of the difficulties in distinguishing pixel values of the input data and due to the unbalanced distribution of its classes.To address these issues, this research introduces a novel approach using a parallel U-Net model for semantic segmentation of permeable surfaces. The process involves binary classification to distinguish between entirely and partially permeable surfaces, followed by fine-grained classification into four distinct permeability levels. Results show that this novel method enhances accuracy and generalizability, particularly when working with small, unbalanced datasets dominated by a single category. Furthermore, this research creates a model capable of generalization across different geographical domains. Domain adaptation is explored to transfer knowledge from one location to another, addressing the challenges posed by varying environmental characteristics. Experiments demonstrate that the parallel U-Net model outperforms the baseline U-Net when applied across domains. To support this research, a permeable surface dataset is introduced, covering three U.S. counties. Because the dataset is comprised of three counties this allows for cross-area performance evaluation. Pixel-wise fine-grained labeling is provided for five distinct permeable surface classes. This dataset is novel, contributing to future research in permeable surface studies. In summary, this research offers an innovative solution to permeable surface mapping, extends the boundaries of arid environment mapping, introduces a large-scale permeable surface dataset, and explores cross-domain application of deep learning models. The three contributions are enhancing the efficiency and accuracy of permeable surface mapping while progressing in this field.
Description
Date
2024-01-01