Publicação
Assessing Berry Number for Grapevine Yield Estimation by Image Analysis: Case Study with the Red Variety “Syrah”
| Resumo: | The yield estimation provides information that help growers to make decisions in order to optimize crop growth and to organize the harvest operations in field and in the cellar. In most vineyard estates yield is forecasted using manual methods. However, image analysis methods, which are less invasive low cost and more representative are now being developed. The main objective of this work was to estimate yield through data obtained in the frame of Vinbot project during the 2019 season. In this thesis, images of the grapevine variety Syrah taken in the laboratory and in the vineyards of the “Instituto Superior de Agronomia” in Lisbon were analyzed. In the laboratory the images were taken manually with an RGB camera, while in the field vines were imaged either manually and by the Vinbot robot. From these images, the number of visible berries were counted with MATLAB. From the laboratory values, the relationships between the number of visible berries and actual bunch weight and berry number were studied. From the data obtained in the field, it was analyzed the visibility of the berries at different levels of defoliation and the relationship between the area of visible bunches and the visible berries. Berry-by-berry occlusion showed a value of 6.4% at pea-size, 14.5% at veraison and 25% at maturation. In addition, high and significant determination coefficient were obtained between actual yield and visible berries. The comparison of estimated yield, obtained using the regression models with actual yield, showed an underestimation at all the three phonological stages. This low accuracy of the developed models show that the use of algorithms based on visible berry number on the images to estimate yield still needs further research |
|---|---|
| Autores principais: | Attinà, Giorgia |
| Assunto: | yield estimation algorithm precision viticulture image analysis VINBOT |
| Ano: | 2021 |
| 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: | The yield estimation provides information that help growers to make decisions in order to optimize crop growth and to organize the harvest operations in field and in the cellar. In most vineyard estates yield is forecasted using manual methods. However, image analysis methods, which are less invasive low cost and more representative are now being developed. The main objective of this work was to estimate yield through data obtained in the frame of Vinbot project during the 2019 season. In this thesis, images of the grapevine variety Syrah taken in the laboratory and in the vineyards of the “Instituto Superior de Agronomia” in Lisbon were analyzed. In the laboratory the images were taken manually with an RGB camera, while in the field vines were imaged either manually and by the Vinbot robot. From these images, the number of visible berries were counted with MATLAB. From the laboratory values, the relationships between the number of visible berries and actual bunch weight and berry number were studied. From the data obtained in the field, it was analyzed the visibility of the berries at different levels of defoliation and the relationship between the area of visible bunches and the visible berries. Berry-by-berry occlusion showed a value of 6.4% at pea-size, 14.5% at veraison and 25% at maturation. In addition, high and significant determination coefficient were obtained between actual yield and visible berries. The comparison of estimated yield, obtained using the regression models with actual yield, showed an underestimation at all the three phonological stages. This low accuracy of the developed models show that the use of algorithms based on visible berry number on the images to estimate yield still needs further research |
|---|