Automated system for detecting and locating agricultural pests in the citrus industry - using radar and IoT -, promoting instant interconnectivity beteween, the fiel, the cloud and the end user for assertive decisions-making.

• Grant number: 24/01106-1
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Start date: January 01, 2025
• End date: December 31, 2026
• Field of knowledge: Agronomical Sciences - Agronomy - Plant Health
• Principal Investigator: Regina Bernardina Johanna Hakvoort
• Grantee: Regina Bernardina Johanna Hakvoort
• Company: Anáhata Soluções Ltda
• CNAE: Cultivo de laranja ; Atividades de apoio à agricultura
• City: Piracicaba
• Associated researchers: Hugo Enrique Hernández Figueroa
• Associated scholarship(s): 25/03438-4 - Development of prediction models using convolutional neural network algorithms for the detection of pests and parasitoids of economic importance in citrus farming., BP.TT; 25/00349-0 - Development of a user interface and geolocation data from the field, linking and integrating it with the neural network database developed by another fellow., BP.TT; 25/02339-2 - Effective development in collecting insect frequencies using radar, developing knowledge in Matlab and Phyton based on biological identification of insect species (pests and parasitoids) and varied stimuli in an anechoic chamber., BP.TT; 24/19248-7 - Automated system for detecting and locating agricultural pests in the citrus industry, using radar and IoT, promoting instant interconnectivity beteween, the fiel, the cloud and the end user for assertive decisions-making, BP.PIPE

Abstract

Automated pest detection using sensors, hardware, neural networks and software is showing timely and promising growth in the agribusiness sector, requiring research and new technological advances in the consumer market.The area of interest will be 100% automated agricultural pest monitoring, with continuous operation (24/7) capable of making 1 to 24 rotations per day in an "illuminated" area with the use of radar, that is, pest detection will occur in time real and continuous operation in citrus fields illuminated by this solution, making it possible to identify behavior patterns of pests present in the location of interest. This practice will cause an exponential increase in the creation of new pest control strategies, both for citrus and other crops of interest to the producer, since site monitoring will be continuous and added to the identification and behavior of the insect and/or pest, will result in decision-making to combat the local pest.Furthermore, the information obtained by the radars from the identification of each species of insect pests and/or parasitoids identified by the equipment in real time and with precision (quantity and coordinates) throughout the cultivated area targeted by the radars will be of extreme importance to assist in decision-making, such as new actions and/or more assertive actions in combating pest(s) of interest for local extermination, ensuring greater crop productivity, using the correct map.Another point that we must draw attention to concerns the fact that the producer and/or crop managers will not need to go to the cultivated area/crop, since the information obtained by radars is precise, automated, easily accessible and intuitive.The study aims to contribute to the state of the art by employing C-band radar technology for the detection of agricultural pests, mainly in cotton, citrus and soybean crops, using the Doppler configuration, with emphasis on citrus, which will be the object of study of the this project. The experiment will involve characterizing several pest species using spectral signatures detected by radar in a controlled environment. Insects will be exposed to leaves from the same crop they infest, increasing the realism of behavior similar to field conditions. Subsequently, a convolutional neural network similar to LeNet 5 will be deployed to train different models for each type of crop, in this case citrus, counting/identifying attacking pests specific to the crop of interest. As an example, we can cite the results obtained in previous research carried out in partnership with EMBRAPII - ESALQ Unit, department of Entomology, coordinated by Prof. Pedro Takao Yamamoto in data validation ranged from 80.92% to 90.24%. The results obtained in the research allowed us to conclude that the robust classification of different pest species is viable, even when their behaviors are different. This approach offers a promising path to timely and accurate pest management, mitigating potential crop losses and contributing to agricultural sustainability.Keywords: radar, various crops, pesticides, pests and diseases. (AU)