A carregar...
Publicação
Vineyards monitoring using convolutional neural networks and multispectral images
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 58.95 MB | Adobe PDF |
Resumo(s)
This study presents the development of an agricultural monitoring system designed to detect
vineyards and crop lines through the application of binary segmentation techniques. The primary
objective is to enhance the efficiency of vineyard monitoring, enabling precise plant detection
using aerial imagery captured by unmanned aerial vehicles (UAVs). The system utilizes U-Net
architecture for semantic segmentation, which was selected for its ability to effectively differentiate between vine and non-vine areas, promoting resource optimization and sustainable viticulture.
Additionally, an algorithm based on the Hough Transform was implemented to accurately detect
vineyard crop rows, further supporting precision agriculture practices. The model was trained
and validated using datasets obtained from various sources, including publicly available datasets
and those provided by industry partners. Evaluation metrics such as accuracy, Intersection over
Union (IoU), and Dice Coefficient were employed to assess model performance, with results indicating varying levels of success across different datasets. The research contributes to the growing
field of precision agriculture by offering a practical tool for vineyard management, with potential applications in resource allocation, environmental sustainability, and operational efficiency.
The system’s design and the methodologies employed underscore the feasibility of integrating advanced machine learning models into real-world agricultural contexts.The code and dataset are
publicly https://github.com/rodrigo-99ferreira/Vineyards
Descrição
Tese de mestrado, Engenharia Informática, 2024, Universidade de Lisboa, Faculdade de Ciências
Palavras-chave
Agricultura de precisão Deteção de vinhas Segmentação binária Arquitetura U-Net Deteção de linhas de cultivo Teses de mestrado - 2024