Publicação
Comparison of Multispectral and Hyperspectral UAV Imagery for Late Blight (Phytophtora infestans) detection in a potato (Solanum tuberosum) field
| Resumo: | Late Blight (LB) is of high concern in the potato crop production. The disease, caused by the oomycete Phytophtora infestans, is responsible for causing huge impacts on the global production. A prompt and specific control, besides being most likely to succeed, enables strategies of small-scale site-specific management which, when combined with a correct detection and good decision model, are expected to reduce the pesticide use. The detection of diseased plants with multispectral and hyperspectral UAV imagery has a great potential in improving LB management and control. In this context, the Portuguese company HPDRONES, within the PG-PSA project, wants to investigate the use of hyperspectral solutions for LB early detection. A drone embedded with a hyperspectral camera has been flighted across an experimental plot in the Ribatejo region. The present work provide an exemplification of a model for the detection of LB using UAV imagery in an open field of potato, involving the use of drones for the image collection and of machine learning algorithms to implement the data analysis. A supervised classification method is proposed. The result is a hyperspectral classifier able to perform an accurate LB detection. On the basis of the limitation of the hyperspectral technologies, a sensitivity approach is proposed, in which spatial and spectral degraded data are used to evaluate the classification’s success under different conditions. The results suggest that the use of drones with multispectral sensors or using different operational parameters (i.e. flight altitude) do not affect the efficiency of the LB detection. |
|---|---|
| Autores principais: | La Parola, Costanza Maria |
| Assunto: | remote sensing disease detection classification methods machine learning deep learning detecão remota deteção de doenças métodos de classificação algoritmos de aprendizagem |
| Ano: | 2022 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório da Universidade de Lisboa |
| Resumo: | Late Blight (LB) is of high concern in the potato crop production. The disease, caused by the oomycete Phytophtora infestans, is responsible for causing huge impacts on the global production. A prompt and specific control, besides being most likely to succeed, enables strategies of small-scale site-specific management which, when combined with a correct detection and good decision model, are expected to reduce the pesticide use. The detection of diseased plants with multispectral and hyperspectral UAV imagery has a great potential in improving LB management and control. In this context, the Portuguese company HPDRONES, within the PG-PSA project, wants to investigate the use of hyperspectral solutions for LB early detection. A drone embedded with a hyperspectral camera has been flighted across an experimental plot in the Ribatejo region. The present work provide an exemplification of a model for the detection of LB using UAV imagery in an open field of potato, involving the use of drones for the image collection and of machine learning algorithms to implement the data analysis. A supervised classification method is proposed. The result is a hyperspectral classifier able to perform an accurate LB detection. On the basis of the limitation of the hyperspectral technologies, a sensitivity approach is proposed, in which spatial and spectral degraded data are used to evaluate the classification’s success under different conditions. The results suggest that the use of drones with multispectral sensors or using different operational parameters (i.e. flight altitude) do not affect the efficiency of the LB detection. |
|---|