Estudo comparativo de métodos geoestatísticos aplicados em agricultura de precisão
Carregando...
Arquivos
Data
2004
Autores
Vicente, Jaqueline [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
O rendimento das culturas pode variar no espaço em decorrência da heterogeneidade das formas do relevo, dos tipos de solo e do manejo ao qual estes solos têm sido submetidos. Como uma de suas estratégias principais, a agricultura de precisão fundamenta-se na coleta sistemática e o processamento de dados georreferenciados para produzir informações que permitam manter os processos produtivos sob controle, orientando-os ao aprimoramento contínuo. Na agricultura de precisão tem sido adotados dois métodos para obter superfícies que apresentam a distribuição espacial do rendimento. O primeiro utiliza amostragem aleatória do rendimento. O segundo faz uso de uma amostragem sistemática com colhedeira equipada com receptor GPS e um dispositivo de estimativa de peso dos grãos colhidos. Este oferece medidas de rendimentos mais precisas, porém a um custo mais elevado. Sabendo-se que imagens multiespectrais de sensoriamento remoto podem apresentar correlação com a vegetação, as mesmas podem ser utilizadas como dados adicionais no processo de reconstrução de superfícies. Nesta pesquisa, realizou-se o estudo comparativo dos métodos de krigagem ordinária isotrópica e anisotrópica, indicadora isotrópica e a cokrigagem, utilizando dois conjuntos de dados de rendimento de soja, além de imagens multiespectrais obtidas com uma câmara digital. Os resultados obtidos no primeiro experimento, mostraram que o processo de krigagem anisotrópica foi o mais adequado na modelagem do rendimento da soja. Para o segundo experimento, os resultados confirmaram que o processo de krigagem ordinária anisotrópica foi o mais adequado. Porém, os índices de concordância kappa para o segundo experimento não mostraram resultados muito satisfatórios na comparação entre o mapa de referência e os mapas obtidos pelos processos de krigagem...
The crop yield can vary in the space due to the heterogeneity of the relief, of the soil types and to the handling to which these soils have been submitted. As one of their main strategies, the precision farming is based on a systematic acquisition and the georreferenced data processing that produce information to allow to maintaing the productive processes under control, guiding them to the continuous improvement. In the project of precision farming it has been adopted two methods to obtain crop yield distribution surfaces. The first adopt random sampling of the income. The second is based on a systematic sampling with harvester equipped with receiver GPS and a device that estimates the weight of the picked grains. This method offers more accurate measures of crop yields, however at a higher cost. Considering that multiespectral images of remote sensing can be correlated correlation with the vegetation, the same ones can be applied as additional data in the process of surfaces reconstruction. In this research, it has been acomplished a comparative study of the methods of isotropic and anisotropic ordinary kriging, isotropic indicator kriging and the cokriging, using two soybean crop yield data sets, besides multiespectral images obtained with a digital camera. The results of the first experiment, for a crop area reasonably big, showed that the process of anisotropic kriging was the most appropriate in the modelling of the crop yield of the soy. For the second experiment, the results confirmed that the process of anisotropic ordinary kriging was the most appropriate. However, the kappa indexes was not very satisfactory. In that last experiment, it was still tried to accomplish the cokriging process using off nadir multiespectral images acquired by the DuncaTech MS-3100 digital camera. However, it was not possible the accomplishment of the cokriging process...(Complete abstract click electronic access below)
The crop yield can vary in the space due to the heterogeneity of the relief, of the soil types and to the handling to which these soils have been submitted. As one of their main strategies, the precision farming is based on a systematic acquisition and the georreferenced data processing that produce information to allow to maintaing the productive processes under control, guiding them to the continuous improvement. In the project of precision farming it has been adopted two methods to obtain crop yield distribution surfaces. The first adopt random sampling of the income. The second is based on a systematic sampling with harvester equipped with receiver GPS and a device that estimates the weight of the picked grains. This method offers more accurate measures of crop yields, however at a higher cost. Considering that multiespectral images of remote sensing can be correlated correlation with the vegetation, the same ones can be applied as additional data in the process of surfaces reconstruction. In this research, it has been acomplished a comparative study of the methods of isotropic and anisotropic ordinary kriging, isotropic indicator kriging and the cokriging, using two soybean crop yield data sets, besides multiespectral images obtained with a digital camera. The results of the first experiment, for a crop area reasonably big, showed that the process of anisotropic kriging was the most appropriate in the modelling of the crop yield of the soy. For the second experiment, the results confirmed that the process of anisotropic ordinary kriging was the most appropriate. However, the kappa indexes was not very satisfactory. In that last experiment, it was still tried to accomplish the cokriging process using off nadir multiespectral images acquired by the DuncaTech MS-3100 digital camera. However, it was not possible the accomplishment of the cokriging process...(Complete abstract click electronic access below)
Descrição
Palavras-chave
Cartografia, Processamento de imagens - Tecnicas digitais, Imagens digitais, Geologia - Métodos estatísticos, Camera digital
Como citar
VICENTE, Jaqueline. Estudo comparativo de métodos geoestatísticos aplicados em agricultura de precisão. 2004. 163 f. Dissertação (mestrado) - Universidade Estadual Paulista, Faculdade de Ciências e Tecnologia, 2004.