| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.39 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
A presença da mosca-da-azeitona (Bactrocera oleae) é uma das principais ameaças do setor olivícola, podendo provocar perdas significativas na qualidade e na produtividade. Perante esta ameaça, esta dissertação procura uma solução viável para a poder monitorizar, com diminuição de custos operacionais e identificação automática. O objetivo prende-se com o desenvolvimento de uma armadilha inteligente de custo acessível que identifique e realize a contagem automática diária do número de moscas. A dissertação foi realizada em várias etapas. 1) Desenvolvimento e construção da armadilha (hardware e software), incluindo testes do seu protótipo em laboratório e no terreno, em olivais de produção; 2) Treino do algoritmo para identificação da mosca-da-azeitona; 3) Teste do algoritmo; 4) Desenvolvimento do dashboard; e 5) Comparação da armadilha inteligente com armadilhas tradicionais. O sistema foi equipado com um modelo de visão computacional, baseado no algoritmo YOLO11 (You Only Look Once), que consegue processar e analisar as imagens capturadas. Os dados recolhidos foram enviados automaticamente para armazenamento em nuvem, que está associado a um dashboard interativo. Este sistema reduz a necessidade de deslocações ao campo, aumenta a precisão da informação e permite uma resposta mais rápida e eficiente no combate desta praga. Sintetizando os resultados desta dissertação, conseguiu-se ter uma armadilha com o design idealizado, o algoritmo desenvolvido apresentou um desempenho satisfatório com uma precisão geral de 0.837, um recall de 0.811 e um mAP50 de 0.86, o dashboard desenvolvido apresenta uma interface simples e intuitiva que permiti visualizar os dados obtidos de forma automática. Em comparação de custos sobre as armadilhas tradicionais, para o exemplo utilizado, representa um investimento elevado (18 298,40€) no primeiro ano, mas que representa nos anos seguintes uma poupança de 2 415€. A solução proposta melhora a eficiência da monitorização da mosca-da-azeitona, otimizando os recursos e minimizando os estragos causados pela praga.
The presence of the olive fruit fly (Bactrocera oleae) is one of the main threats to the olive sector, potentially causing significant losses in quality and productivity. To address this threat, this dissertation seeks a viable solution for monitoring it, reducing operational costs, and enabling automatic identification. The objective is to develop an affordable smart trap that identifies and automatically counts flies daily. The dissertation was carried out in several stages: 1) Development and construction of the trap (hardware and software), including testing of its prototype in the laboratory and in the field in production olive groves; 2) Training the algorithm for olive fruit fly identification; 3) Testing the algorithm; 4) Development of the dashboard; and 5) Comparison of the smart trap with traditional traps. The system was equipped with a computer vision model, based on the YOLO11 (You Only Look Once) algorithm, which can process and analyze the captured images. The collected data was automatically sent to cloud storage, which is linked to an interactive dashboard. This system reduces the need for field trips, increases information accuracy, and enables a faster and more efficient response to combat this pest. Summarizing the results of this dissertation, a trap with the ideal design was created. The developed algorithm demonstrated satisfactory performance, with an overall precision of 0.837, a recall of 0.811, and a mAP50 of 0.86. The dashboard features a simple and intuitive interface that allows for automatic visualization of the data obtained. In comparison to the costs of traditional traps, for the example used, this represents a high investment (€18,298.40) in the first year, but represents a savings of €2,415 in subsequent years. The proposed solution improves the efficiency of olive fly monitoring, optimizing resources and minimizing damage caused by the pest.
The presence of the olive fruit fly (Bactrocera oleae) is one of the main threats to the olive sector, potentially causing significant losses in quality and productivity. To address this threat, this dissertation seeks a viable solution for monitoring it, reducing operational costs, and enabling automatic identification. The objective is to develop an affordable smart trap that identifies and automatically counts flies daily. The dissertation was carried out in several stages: 1) Development and construction of the trap (hardware and software), including testing of its prototype in the laboratory and in the field in production olive groves; 2) Training the algorithm for olive fruit fly identification; 3) Testing the algorithm; 4) Development of the dashboard; and 5) Comparison of the smart trap with traditional traps. The system was equipped with a computer vision model, based on the YOLO11 (You Only Look Once) algorithm, which can process and analyze the captured images. The collected data was automatically sent to cloud storage, which is linked to an interactive dashboard. This system reduces the need for field trips, increases information accuracy, and enables a faster and more efficient response to combat this pest. Summarizing the results of this dissertation, a trap with the ideal design was created. The developed algorithm demonstrated satisfactory performance, with an overall precision of 0.837, a recall of 0.811, and a mAP50 of 0.86. The dashboard features a simple and intuitive interface that allows for automatic visualization of the data obtained. In comparison to the costs of traditional traps, for the example used, this represents a high investment (€18,298.40) in the first year, but represents a savings of €2,415 in subsequent years. The proposed solution improves the efficiency of olive fly monitoring, optimizing resources and minimizing damage caused by the pest.
Descrição
Dissertação de Mestrado, Ciência de Dados em Agricultura, Alimentação, Floresta e Ambiente, 2025, Universidade de Lisboa, Instituto Superior de Agronomia.
Palavras-chave
armadilha inteligente mosca-da-azeitona YOLO11 deteção automática olivicultura smart trap olive fly automatic detection olive farming
