Embedded system for monitoring coffee productivity, maturation and anomalies

• Grant number: 19/23048-5
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Duration: March 01, 2021 - February 28, 2023
• Field of knowledge: Agronomical Sciences - Agricultural Engineering
• Principal Investigator: Jose Angelo Gurzoni Junior
• Grantee: Jose Angelo Gurzoni Junior
• Host Company: Adroit Sistemas Inteligentes, Equipamentos para Automação, Consultoria e Serviços de Engenharia Ltda
• CNAE: Cultivo de café
• City: São Paulo
• Associated researchers: Abner Matheus Costa de Araújo ; Caio César Teodoro Mendes ; Cristiane Silva Ferreira ; José Guilherme Mota Esgario ; Milton Peres Cortez Junior ; Plinio Thomaz Aquino Junior ; Samuel Ribeiro Giordano
• Associated scholarship(s): 22/00749-0 - Fruits detection and classification with temporal evolution mapping applied in coffee plantations, BP.TT; 22/01651-4 - Automatic detection of anomalies in leaves and fruits in coffee plantations, BP.TT

Abstract

With extensive research and development of in-field smart sensors, Adroit Robotics has worked during the last few years on the precise monitoring of productivity indexes and health status of fruit crops, as well as early detection of diseases. This technology, entering operational stage in citrus orchards, has great potential of application to coffee orchards, in which the monitoring of the fruit maturing levels has the potential to vastly increase the productivity of superior quality beans and to contain the spread of important diseases and pests like the coffee rust, coffee leaf miner and the coffee berry borer. The early contention of diseases is important and not only because of its large financial impact, but also because it allows the reduction of the quantity of agrochemicals and, thus, the reduction of risks to the environment and live beings. This research project has the purpose of applying image processing and machine learning techniques to the detection of fruits and leaves for monitoring of productivity, maturing stage and early detection of diseases in coffee orchards. The project considers that the detection equipment and apparatus will be mounted on tractors and agricultural machinery already in use for the crop management. The techniques and methods proposed were already applied by the company in citrus orchards of large producers in the State of Sao Paulo, with good results.Embedding the system in machinery already in use allows the optimization of resources for large scale data collection without added tasks and costs to the producers. The system will use very high-resolution imagery to collect fruits quantity, distribution, ripe stage, geolocation, tree heights and spacing, tree volumes, environmental conditions and other variables that may be required for the monitoring of the crop yield and health. The use of an automated device, without any human intervention, allows for precise, large scale and continuous monitoring of the orchard, a breakthrough when compared to the small samplings used for decision making with traditional methods.The proposed system can also be used for the correct crop yield estimation. The crop yield estimation is currently made with harvesting of about 1% of the trees, which cannot account for the variability of the crop and results in difficulties to produce detailed estimates. Another challenge for coffee producers is the definition of the optimal harvest schedule, which directly affects the crop quality and its value. Early harvesting leads to too many green beans in the mix, increasing the acidity of the drink, whereas late harvesting leads to black fruits, which also reduce the quality of the drink and can increase pest proliferation throughout the orchard. Several pests affect the development and productivity of the coffee, causing direct economic losses. There are, of course, preventive methods for pest control, like the application of agrochemicals in advance during season stages or climates where these pests typically occur. However, such practices lead to waste and can cause unbalances that favor the occurrence of other pests. The early detection of disease and pest spots leads to rationalization of agrochemical utilization.The scientific and technical challenges to be overcome during the execution of this proposal involve the development of the sensor hardware, the research of the appropriate computer vision and machine learning algorithms for the estimates of productivity, maturing and anomalies in coffee orchards, as well as the development of high-performance software using these algorithms. At the end of this PIPE phase 2 (direct) project we expect to obtain consistent results proving the efficiency of the detection of fruit, leaves and diseases using optical sensors, computer vision and machine learning in an automated data collection process. We also expect to confirm its economic feasibility for commercial scale operation. (AU)