Millora i validació d'un model basat en imatges satèl·lit per caracteritzar vegetació en vinya
An effective and efficient characterization of the vegetation is key to being able to determine the appropriate amount of phytosanitary product to be applied when carrying out phytosanitary treatments in the vineyard. Vegetation evolves and develops based on factors linked to the soil, the climate o...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis de maestría |
| Fecha de publicación: | 2022 |
| País: | España |
| Institución: | Universitat Politècnica de Catalunya (UPC) |
| Repositorio: | UPCommons. Portal del coneixement obert de la UPC |
| Idioma: | catalán |
| OAI Identifier: | oai:upcommons.upc.edu:2117/376356 |
| Acceso en línea: | https://hdl.handle.net/2117/376356 |
| Access Level: | acceso abierto |
| Palabra clave: | Agriculture--Remote sensing Índexs de vegetació Caracterització Vinya Estadis fenològics Satèl·lits Sensors remots Tècniques agrícoles Àrees temàtiques de la UPC::Enginyeria de la telecomunicació::Radiocomunicació i exploració electromagnètica::Teledetecció |
| Sumario: | An effective and efficient characterization of the vegetation is key to being able to determine the appropriate amount of phytosanitary product to be applied when carrying out phytosanitary treatments in the vineyard. Vegetation evolves and develops based on factors linked to the soil, the climate or the physiology of the plant itself, and that is why within the same plot there is a certain spatial variability in crop growth. It has been proven that expressing the application volumes based on parameters linked to the crop, such as the leaf volume per surface unit, provide significant savings and improvements in the quality of the applications. These new methods of dose expression are based on manual measurements in the field, which are costly to carry out and sometimes slow down the introduction of these methodologies. New technologies such as remote sensing provide information at a global level and local resolutions, together with a high frequency of data. It is for this reason that an opportunity opens up to automate data acquisition processes, which in the case of the application of phytosanitary products involves being able to estimate the characteristics of the crop automatically. The aim of this project was to improve an existing mathematical model that allows to relate vegetation indices calculated from satellite images with vegetation measurements in different vegetative stages in vineyard cultivation. The existing model includes data from the years 2018 and 2019 and has a single vegetation index as an independent variable. During this project, the model was complemented with data acquired during the years 2021 and 2022, and 7 new spectral indices were calculated that served to develop new prediction models. The results showed that the existing model based on the Normalized Differential Vegetation Index (NDVI) and used to estimate the width of the vegetation, lost its predictive capacity when adding samples from 2021 and 2022. Similarly, the Normalized Red Edge Vegetation Index (NDRE) appeared as a potential candidate to estimate the height and width of the vegetation with determination coefficients above 0,55 and 0,45, respectively. When data were grouped by vigor class, all measured spectral indices were significantly correlated with crop height, with coefficients of determination above 0,55. With these results, it can be concluded that satellite images are a promising tool for the remote characterization of vineyard cultivation, although due to their technical limitations, in early phenological stages, with shoots below 40 cm, it will be difficult to obtain data reliable. Once the vineyard is developed, both the NDVI and the NDRE are indices that can provide us with good results that will help implement new management strategies for phytosanitary products based on variable application technologies. |
|---|