Técnicas avançadas de monitorização em aplicações de agricultura de precisão
Data
2022-09-30
Título da revista
ISSN da revista
Título do Volume
Editora
Resumo
A agricultura perspetiva, a curto, médio e longo prazos, desafios
multidimensionais relacionados com a escassez de recursos, com cenários de
alterações climáticas severas e com a necessidade de implementar práticas ambiental
e economicamente sustentáveis. Neste contexto de transições ecológica, energética
e digital, ter-se-á de desenvolver e implementar soluções disruptivas, de forma a
garantir a tão necessária produção de alimento. Incorporar tecnologia da indústria,
antever e debelar problemas de forma integrada, otimizar recursos e processos
produtivos, produzir de forma diferenciada e recorrer a ferramentas tecnológicas
inovadoras, que incorporem inteligência artificial, é o mote e faz parte da agenda
global para a próxima década.
A viticultura vive um cenário em tudo semelhante. No caso da Região Demarcada do
Douro, acrescem, ainda, os problemas de falta de recursos humanos para diferentes
tarefas essenciais na cadeia de produção. A disseminação e a implementação de
práticas de vitivinicultura de precisão, geridas por sistemas fortemente apoiados
em dados que emergem de várias fontes e se apresentam em diversos formatos,
tem vindo a crescer. Para além de séries temporais de dados agro-meteorológicos,
particularmente relevantes em vários tipos de previsões, as imagens aéreas têm
sobressaído como um auxiliar de enorme potencial na gestão de recursos, no
planeamento, na construção de mapas de prescrição e, até mesmo, das próprias
explorações. A evolução tecnológica dos sistemas embebidos, da capacidade das
telecomunicações e o baixo custo inerente potenciou novas formas de monitorizar
processos e de processar dados de forma distribuída. A incorporação de inteligência
artificial tornou essa monitorização capaz de extrair informação de forma mais rápida
e mais eficiente, numa ótica embrionária de Data Insight, onde o conhecimento tem
como base um Data Lake.
Este trabalho de investigação aborda o uso de inteligência artificial na monitorização
inteligente e na classificação automática, enquadradas na viticultura de precisão,
com o objetivo de criar soluções disruptivas e inovadoras que permitam acrescentar
valor ao setor da vinha e do vinho. Para além de ser apresentada uma plataforma
inteligente baseada no paradigma IoT (Internet of Things) – mySense - , capaz
de dar resposta automática a solicitações, foi desenvolvido, também, o sistema
VineInspector. Trata-se de um dispositivo embebido IoT de baixo custo e de fácil
instalação na vinha, que recorre a sensores de imagem e a inteligência artificial para
executar tarefas de particular relevância nas gestão da vinha, como seja a contagem
automática de pragas em armadilhas, o seguimento automático da fenologia da
videira, e a previsão de infeção por míldio.
Os resultados alcançados a partir de ensaios realizados ao longo de dois anos
em várias localizações e em condições reais de utilização, mostram as claras
vantagens que o VineInspector oferece neste tipo de práticas. Destaca-se o registo
de informação visual relevante e a emissão de notificações automáticas, sem a
necessidade de deslocação ao terreno. Para além deste dispositivo, que se constitui
como o corolário deste trabalho, muitas outras contribuições ficam disponibilizadas
como serviços mySense. Por exemplo, a classificação automática de castas de videira
com base em imagem adquirida em ambiente real.
Embora com muitas perspetivas de trabalho futuro, poder-se-á concluir que este
trabalho deixa um conjunto de processos e de aplicações iniciados para que
seja possível uma desejável transferência tecnológica para empresas do setor que
pretendam abraçar este tipo de soluções e as tornem disponíveis. É importante
assinalar que este trabalho constituiu a base que permitiu criar uma empresa
tecnológica - a Kropie - que se espera possa vir a desempenhar um papel
fundamental nos processos de inovação contínua e de transferência de tecnologia
no sector da vinha e do vinho, quer nacional, quer internacional.
In the short, medium and long term, agriculture faces multidimensional challenges related to the scarcity of resources, severe climate change scenarios and the need to implement environmentally and economically sustainable practices. In this context of ecological, energy and digital transitions, disruptive solutions will have to be developed and implemented in order to guarantee muchneeded food production. Incorporating industry technology, anticipating and solving problems in an integrated way, optimising resources and production processes, producing differently and using innovative technological tools that incorporate artificial intelligence is the goal and is part of the global agenda for the next decade. Viticulture lives a very similar scenario. In the case of the Douro Demarcated Region, additional problems are the lack of human resources for different essential tasks in the production chain. The dissemination and implementation of precision viticulture practices, managed by systems strongly supported by data that emerge from various sources and are presented in various formats, has been growing. In addition to time series of agro-meteorological data, particularly relevant in various types of forecasting, aerial images have stood out as an aid of enormous potential in resource management, planning, prescription mapping and even of the vineyards themselves. The technological evolution of embedded systems, the capacity of telecommunications and the inherent low cost have enabled new ways of monitoring processes and processing data in a distributed manner. The incorporation of artificial intelligence has made this monitoring capable of extracting information faster and more efficiently, in an embryonic perspective of "Data Insight", where knowledge is supported by a "Data Lake". This research work addresses the use of artificial intelligence in intelligent monitoring and automatic classification, within the context of precision viticulture, with the aim of creating disruptive and innovative solutions that add value to the vineyard and wine sector. In addition to presenting an intelligent platform based on the IoT (Internet of Things) paradigma – mySense - capable of responding automatically to solicitations, the VineInspector system was also developed. This is a lowcost IoT embedded device that is easy to install in the vineyard and uses image sensors and artificial intelligence to perform tasks of particular relevance in vineyard management, such as the automatic counting of trapped plagues, the automatic monitoring of the phenology of the vine, and the prediction of infection by downy mildew. The results obtained from tests carried out over two years in several vineyards and under real conditions of use have shown the clear advantages that VineInspector offers in this type of practice. It registers relevant visual information and sends automatic notifications without the need to go into the field. Besides this device, which is the corollary of this work, many other contributions are made available as mySense services. For example, the automatic classification of grapevine varieties based on images acquired in a real environment. Although with many perspectives for future work, it can be concluded that this work has initiated a set of processes and applications to enable a desirable technological transfer to companies in the sector that wish to embrace this type of solutions and make them available. It is important to point out that this work was the basis that allowed the creation of a technological company – Kropie - which is expected to play a key role in the processes of continuous innovation and technology transfer in the vineyard and wine sector, both nationally and internationally.
In the short, medium and long term, agriculture faces multidimensional challenges related to the scarcity of resources, severe climate change scenarios and the need to implement environmentally and economically sustainable practices. In this context of ecological, energy and digital transitions, disruptive solutions will have to be developed and implemented in order to guarantee muchneeded food production. Incorporating industry technology, anticipating and solving problems in an integrated way, optimising resources and production processes, producing differently and using innovative technological tools that incorporate artificial intelligence is the goal and is part of the global agenda for the next decade. Viticulture lives a very similar scenario. In the case of the Douro Demarcated Region, additional problems are the lack of human resources for different essential tasks in the production chain. The dissemination and implementation of precision viticulture practices, managed by systems strongly supported by data that emerge from various sources and are presented in various formats, has been growing. In addition to time series of agro-meteorological data, particularly relevant in various types of forecasting, aerial images have stood out as an aid of enormous potential in resource management, planning, prescription mapping and even of the vineyards themselves. The technological evolution of embedded systems, the capacity of telecommunications and the inherent low cost have enabled new ways of monitoring processes and processing data in a distributed manner. The incorporation of artificial intelligence has made this monitoring capable of extracting information faster and more efficiently, in an embryonic perspective of "Data Insight", where knowledge is supported by a "Data Lake". This research work addresses the use of artificial intelligence in intelligent monitoring and automatic classification, within the context of precision viticulture, with the aim of creating disruptive and innovative solutions that add value to the vineyard and wine sector. In addition to presenting an intelligent platform based on the IoT (Internet of Things) paradigma – mySense - capable of responding automatically to solicitations, the VineInspector system was also developed. This is a lowcost IoT embedded device that is easy to install in the vineyard and uses image sensors and artificial intelligence to perform tasks of particular relevance in vineyard management, such as the automatic counting of trapped plagues, the automatic monitoring of the phenology of the vine, and the prediction of infection by downy mildew. The results obtained from tests carried out over two years in several vineyards and under real conditions of use have shown the clear advantages that VineInspector offers in this type of practice. It registers relevant visual information and sends automatic notifications without the need to go into the field. Besides this device, which is the corollary of this work, many other contributions are made available as mySense services. For example, the automatic classification of grapevine varieties based on images acquired in a real environment. Although with many perspectives for future work, it can be concluded that this work has initiated a set of processes and applications to enable a desirable technological transfer to companies in the sector that wish to embrace this type of solutions and make them available. It is important to point out that this work was the basis that allowed the creation of a technological company – Kropie - which is expected to play a key role in the processes of continuous innovation and technology transfer in the vineyard and wine sector, both nationally and internationally.
Descrição
Tese submetida à
UNIVERSIDADE DE TRÁS-OS-MONTES E ALTO DOURO
para obtenção do grau de
DOUTOR
em Engenharia Electrotécnica e de Computadores, de acordo com o disposto no
DR - I série-A, Decreto-Lei nº 74/2006 de 24 de Março e no
Regulamento de Estudos Pós-Graduados da UTAD
DR, 2ª série - Deliberação nº 2391/2007
Palavras-chave
Viticultura de precisão , Inteligência artificial