Biblioteca do Café
URI permanente desta comunidadehttps://thoth.dti.ufv.br/handle/123456789/1
Navegar
3 resultados
Resultados da Pesquisa
Item Remotely piloted aircraft and computer vision applied to coffee growing management(Universidade Federal de Lavras, 2022-11-25) Santana, Lucas Santos; Ferraz, Gabriel Araújo e SilvaDigital and precision agriculture technologies used in coffee farming have gained space and have become necessary in many coffee production stages. Among the emerging technologies, the Remotely Piloted Aircraft (RPA) can be highlighted because their products can be used as data providers for machine learning techniques and automated monitoring forms. This study aimed to apply cartographic and photogrammetric products from RPAs submitted to machine learning techniques and image analysis in digital and precision coffee farming. Three types of research were built: Application of RPA cartographic products for the coffee plant implantation project; Identification and counting of plants in PRA images and Investigations of plants development in renewal areas. (I)The first study evaluated different flight mission composition efficiency and point cloud levels for Digital Terrain Models generation applied in coffee plantations. Flights performed at 120 m Above Ground Land (AGL) and 80 × 80% overlap showed higher assertiveness and efficiency. The 90 m AGL flight showed great terrain detail, causing significant surface differences concerning the topography obtained by Global Navigation Satellite System (GNSS) receivers. Slope ranges up to 20% are considered reliable for precision coffee growing projects. Changes in flight settings and image processing are satisfactory for precision coffee projects. Image overlap reduction significantly lowed the processing time without influencing Digital Terrain Model DTM's quality. (II) The second research aimed to develop an algorithm for automatic counting coffee plants and define the plant's best age to carry the monitoring using RPA images. Plants with four months of development showed 86.5% count assertiveness. The best results were observed in plantations with six months of development, presenting an average of 96.8% of assertiveness in automatically counting plants. This analysis enables an algorithm development for automated counting of coffee plants through RGB images obtained by remotely piloted aircraft and machine learning applications. (III) The objective of the third research was to monitor the coffee plants' development planted on ash from crop residues through vegetative indices in RPA images, analysis of chemical elements presents in the ash and soil analysis. Preliminary results indicate the high presence of aluminum and potassium in the ash, causing significant differences in coffee development beginning. In addition, variations were observed in vegetative indices values in regions with ash presence, highlighting the NGI and NNIRI indices. The research developed by this paper provides essential information for digital agriculture technologies advancement in coffee growing.Item Precision coffee growing: a review(Universidade Federal de Lavras, 2022-06-09) Santana, Lucas Santos; Ferraz, Gabriel Araújo e Silva; Santos, Sthéfany Airane dos; Dias, Jessica Ellen LimaPrecision Agriculture (PA) technologies introduction in coffee-growing is becoming essential to advances in sustainable cultivation and increase in output. Applications that involve PA techniques in coffee production are defined now as Precision Coffee growing (PC). Systematically explored, studies on the subject contribute to improvements in the area, relating soil variability to its impacts on plants. The PC’s scientific approach offers new forms of manage-ment and more security in coffee production. Aimed at reducing pesticides application and nutrients to the soil, contributing to sustainable development in coffee production. Initially, the research on coffee production had dealt with soil spatial variability, highlighting the geostatistical methods and specific ways to sample the soil. With technological advances in agriculture, new ways of monitoring spatial variability are available. In this context, studies are arising on spatial variability related to the plant, applying terrestrial, aerial and orbital sensors, possibly creating perspectives for monitoring and mapping coffee production. Artificial intelligence, Remotely Piloted Aircraft (ARP) products, harvesting yield sensors, automatic grain classifiers, and remote sensing stand out as new technologies under development in coffee production. These applications in PC involving multidisciplinary research demonstrate new relevant ways of improving crop managing and sustainability guaranteeing.Item Treinamento de algoritmo para identificação de mudas de café por meio de imagens aéreas(Embrapa Café, 2019-10) Santana, Lucas Santos; Santos, Gabriel Henrique Ribeiro dos; Ferraz, Gabriel Araújo e Silva; Santos, Luana Mendes dos; Barbosa, Brenon Diennevan Souza; Bento, Nicole LopesAplicação de sistemas computacionais na agricultura vem se tornando necessidade, para incrementos de produtividade. A cafeicultura é considerada um dos mais importantes ramos do agronegócio brasileiro, mas alguns pontos ainda podem ser melhorados. A contagem de mudas de café é feita manualmente a campo, considerada uma técnica que apresenta baixa agilidade principalmente em grandes plantações, causando perda de tempo e desconforto à pessoa encarregada por este trabalho. A utilização de tecnologias avançadas no campo surge no contexto da agricultura de precisão. Esta pesquisa tem como intuito apresentar treinamento de um algoritmo para contagem automática de mudas de cafeeiros a partir de imagens obtidas por veículo aéreo não tripulado (VANT). O algoritmo desenvolvido é baseado numa rede neural artificial do tipo convolucional (RNC). A pesquisa foi realizada em uma área experimental de café em 2 estágios 3 e 6 totalizando 0,4 hectares no município de Bom Sucesso MG. Para a coleta das imagens foi utilizado um veículo aéreo não tripulado (VANT) modelo phanton 4, com capacidade de coletar imagens na composição Red Blue e Green (RGB). Os voos foram feitos com 30 metros de altura, velocidade de 3 m/s, sobreposição frontal e lateral de 80%. Para a criação do ortomosaico as imagens foram processadas no software Agisoft Photoscan e posteriormente tratadas no software Arcgis 10.2. Em seguida foi submetido a fase de treinamento, para a detecção automática das mudas no ortomosaico. Primeiramente foi necessário a criação de um banco de dados, com a finalidade de treinar e validar a rede neural YOLO (You Only Look Once). Após o treinamento foi utilizada a técnica de usar pequenos recortes de uma imagem se chama janelamento, na qual foram realizados 762 recortes. As imagens usadas para o treino da rede neural são distintas das imagens usadas para validação para assim o software demostrar o aprendizado. Os resultados apresentados em plantas de 3 meses não foram satisfatórios, pois ocorreu confusão na identificação entre mudas e plantas daninhas e sombras. Nas imagens de 6 meses foi observado melhor desempenho na identificação por meio de treinamento de algoritmos. A partir do treinamento de algoritmo por meio de aprendizado foi possível detectar mudas de cafeeiros a campo. Portanto para um refinamento e maior precisão do algoritmo é necessário o treinamento em diversas áreas e cultivares diferentes.