Enabling Insecticide Spot Application on Boom Sprayer by Developing Machine Vision and Communcation Components

Despite its effectiveness in improving yield, uniform spraying has negative impacts on the environment and the use of resources. In this research project, we plan to contribute to the adaptation of spot spraying by developing electronic and mechanical components that allow boom sprayers of different size scales to adopt this technology. In this paper, we demonstrate the development of machine vision system that uses artificial intelligence to detect Colorado potato beetles based on YOLO v5 models, and the development of a communication to scheme to integrate the machine vision system with an open-source spraying mechanism via Controller Area Network. Using field images taken on-the-go from a camera mounted on a sprayer, the accuracy of detecting beetles reached 64%, which reveals a promising result given the possibility of enhancing image quality and the computation techniques moving forward. Also, the machine vision node, developed to integrate a machine vision system that consist of two neighbouring cameras, was able to flawlessly translate the detection results to a spraying system mounted on boom sprayer to control 6 nozzles individually. These results allow for scaling up the detection and spraying mechanisms to cover full booms which can reach to widths of 36 m. Beside building on the current developments on the detection and communication components, our next steps will include allowing for real-time spraying by introducing modifications on the pumping and nozzle components.