Click on “Download PDF” for the PDF version or on the title for the HTML version.
If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.
Smart citrus tree sprayer using sensor fusion and artificial intelligence
Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org
Citation: 2021 ASABE Annual International Virtual Meeting 2100525.(doi:10.13031/aim.202100525)
Authors: Victor Partel, Lucas Costa, Yiannis Ampatzidis
Keywords: artificial intelligence, sensor fusion, smart machinery, spraying.
Abstract. Efficient chemical spray applications on crops are vital to reduce off-target drift, economic losses, and negative environmental impacts. Smart sprayer systems use sensors to adjust its application rate in real time based on data collected. In this study, a novel system comprising hardware and software was developed for data collection, analysis, and control of a 500-gallon rear sprayer for citrus trees. This low-cost system was equipped with two RGB cameras and a 2D lidar for acquiring information on tree condition and height, while also collecting GPS information for speed to accurately adjust the spraying pattern for each individual tree. The smart sprayer controlled electronic valves on each nozzle with the feedback of a flow meter. An embedded computer (NVIDIA Xavier NX) was used for data processing, control of the system, and to provide a graphical user interface for user manual input and feedback. The accurate reading of tree height from the lidar was enhanced by the camera images and artificial intelligence (YOLO V3), which can classify citrus trees (e.g., alive or dead) and differentiate citrus trees from other objects found in a grove that does not require spraying (e.g., wind-break trees, posts and construction). Experiments were conducted in experimental fields and commercial groves to analyze the reduction in chemical usage while maintaining effectiveness (32% reduction) and precision in tree canopy height assessment (6% error). The developed system provides a base for improvement in data collection, processing, and control. The results show an improvement on commercially available non-smart spraying systems. The error in tree height can be minimized by optimizing placement of the sensors. The sensor fusion can also support tree health classification to better adjust the spraying, input of variable rate application maps, fruit count, and fruit size estimation.
(Download PDF) (Export to EndNotes)