Publicação
TerraSenseTK: a toolkit for remote soil nutrient estimation
| Resumo: | Intensive farming endangers soil quality in various ways. Researchers show that if these practices continue, humanity will be faced with food production issues. For this matter, Earth Observation, more concretely Soil Sensing, along with Machine Learning, can be employed to monitor several indicators of soil degradation, such as soil salinity, soil heavy metal contamination and soil nutrients estimation. More concretely, Soil Nutrients are of great importance. For instance, to understand which crop better suits the land, the soil nutrients must be identified. However, sampling soil is a laborous and expensive task, which can be leveraged by Remote Sensing and Machine Learning. Several studies have already been developed in this matter, although many gaps still exist. Among them, the lack of cross-dataset evaluations of existing algorithms, and also the steep learning curve to the Earth Observation domain that prevents many researchers from embracing this field. In this sense, we propose TerraSense ToolKit (TSTK), a python toolkit that addresses these challenges. In this work, the possibility to use Remote sensing along with Machine Learning algorithms to per form Soil Nutrient Estimation is explored, additionally, a nutrient estimation toolkit is proposed, and the effectivity of it is tested in a soil nutrient estimation case study. This toolkit is capable of simplifying Remote Sensing experiments and aims at reducing the barrier to entry to the field of Earth Observation. It comes with a preconfigured case study which implements a soil sensing pipeline. To evaluate the usability of the toolkit, experiments with five different crops were executed, namely with Wheat, Barley, Maize, Sunflower and Vineyards. This case study gave visibility to an underlying unbalanced data problem, which is not well addressed in the current State of the Art. |
|---|---|
| Autores principais: | Pereira, Manuel Afonso Soares |
| Assunto: | Deteção remota Estimação de nutrientes no solo Toolkit Imagens por satélite Deteção do solo Sensoriamento do solo Remote sensing Soil nutrient estimation Python Satellite imagery Soil sensing Informatics Engineering . Faculdade de Ciências Exatas e da Engenharia |
| Ano: | 2022 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade da Madeira |
| Idioma: | inglês |
| Origem: | DigitUMa - Repositório da Universidade da Madeira |
| Resumo: | Intensive farming endangers soil quality in various ways. Researchers show that if these practices continue, humanity will be faced with food production issues. For this matter, Earth Observation, more concretely Soil Sensing, along with Machine Learning, can be employed to monitor several indicators of soil degradation, such as soil salinity, soil heavy metal contamination and soil nutrients estimation. More concretely, Soil Nutrients are of great importance. For instance, to understand which crop better suits the land, the soil nutrients must be identified. However, sampling soil is a laborous and expensive task, which can be leveraged by Remote Sensing and Machine Learning. Several studies have already been developed in this matter, although many gaps still exist. Among them, the lack of cross-dataset evaluations of existing algorithms, and also the steep learning curve to the Earth Observation domain that prevents many researchers from embracing this field. In this sense, we propose TerraSense ToolKit (TSTK), a python toolkit that addresses these challenges. In this work, the possibility to use Remote sensing along with Machine Learning algorithms to per form Soil Nutrient Estimation is explored, additionally, a nutrient estimation toolkit is proposed, and the effectivity of it is tested in a soil nutrient estimation case study. This toolkit is capable of simplifying Remote Sensing experiments and aims at reducing the barrier to entry to the field of Earth Observation. It comes with a preconfigured case study which implements a soil sensing pipeline. To evaluate the usability of the toolkit, experiments with five different crops were executed, namely with Wheat, Barley, Maize, Sunflower and Vineyards. This case study gave visibility to an underlying unbalanced data problem, which is not well addressed in the current State of the Art. |
|---|