Crop detection and individual nozzle simulation and activation using monocular vision

  • Henrique Gonçalves Andrade Núcleo de Inovação em Máquinas e Equipamentos Agrícolas, Universidade Federal de Pelotas, RS
  • Gustavo Henrique do Nascimento Centro de Ciências Computacionais, Universidade Federal do Rio Grande, RS
  • Paulo Lilles Jorge Drews Jr Centro de Ciências Computacionais, Universidade Federal do Rio Grande, RS
Keywords: Perception and sensing, Computer vision, Agricultural robotics, Agricultural automation, Embedded system


This work brings the development of a perception system to crop detection and singlenozzle activation and simulation for a pesticide boom sprayer. The purpose is to embed thesystem on the agricultural machine to avoid the waste of pesticide by applying it only where isneeded, and therefore be less aggressive with the soil. The motivation was to use an advancedtechnology implemented simply, capable of being adopted by family farmers, with low cost andefficiency for the problem. Due to the proposal?s motivation to be viable for family farmers,there is a hardware cost limitation, such as the use of a single camera. The method adopted toimplement this is the use of simple hardware running computer vision software that implementscrop detection techniques and a trigger logic for the nozzles. The results showed that the proposalcan identify the crops and activate the nozzles individually, however, the processing time testshows that the algorithm needs an implementation improvement.