Métodos de machine learning aplicados à classificação do uso e ocupação do solo na microbacia do Lago Igapó na cidade de Londrina/Pr
The growth of urban areas is a worldwide phenomenon and its problems have generated immense challenges for society. One of these challenges is related to the drainage of rainwater in an urban environment. With urbanization, the original land cover is replaced by elements that prevent or hinder the i...
| Autor principal: | Duarte, Priscila Gabriela da Silva |
|---|---|
| Formato: | Dissertação |
| Idioma: | Português |
| Publicado em: |
Universidade Tecnológica Federal do Paraná
2022
|
| Assuntos: | |
| Acesso em linha: |
http://repositorio.utfpr.edu.br/jspui/handle/1/30022 |
| Tags: |
Adicionar Tag
Sem tags, seja o primeiro a adicionar uma tag!
|
| Resumo: |
The growth of urban areas is a worldwide phenomenon and its problems have generated immense challenges for society. One of these challenges is related to the drainage of rainwater in an urban environment. With urbanization, the original land cover is replaced by elements that prevent or hinder the infiltration of rainwater. Nature based solutions have proved to be a viable alternative for dealing with problems associated with urban drainage. However, the success of interventions is highly dependent on an understanding of the elements that predominate on the urban surface. In this sense, the main objective of this work was the application of machine learning algorithms in high resolution satellite images to classify land use and occupation in the drainage area of the Igapó lake system, a set of urban lakes in the city of Londrina PR. Images of the Pléiades satellite constellation, with a high level of spatial resolution, were used for the classification. Twelve specific thematic classes were established for classifying land cover. Supervised classification was applied, and the following Machine Learning algorithms were evaluated: Decision Tree (DT), Randon Forest (RT), Support Vector Machine (SVM), K Nearest Neighbors (KNN) and Normal Bayes. DT was the classifier that presented the best performance, both for the global classification and for the individual classes. The values for the Kappa, Precision, Recall, and F1 Score indices were between 90% and 100% for the DT classifier. The green areas, represented by Trees/bushes and Grass, together with roofs, represent the majority classes, with 24% and 23% of the land cover of the study area, respectively. Asphalt and pavement, with 19% of the area, and exposed soil, with 14%, were also classes with expressive participation. The remaining fraction consists of water, including pool surface, and shaded areas. With the refinement of the classes, the spatial resolution, and the quality of the developed mapping, the results presented in this work can be a very useful tool for the elaboration of intervention projects that promote the use of nature based solutions to solve urban drainage problems. |
|---|