Precision Weed Spraying UAV

University of Auckland Final Year Project

February - November 2017

Objective: The aim was to develop a system that autonomously tracks and sprays invasive weeds from a quadcopter UAV platform. The intent is that the research undertaken will, in the years to come, contribute towards a fully autonomous platform that can track and spray individual weeds. This reduces the need for large, expensive spraying rigs that consume vast quantities of harsh chemicals, which have significant economic and environmental impacts.

Contribution: A vision system was utilized to track the motion of an invasive weed relative to the motion of the UAV. The weed selected was the Californian thistle, the most common weed on New Zealand farms. The vision system was integrated with a gimbal tracking mechanism to track the weed in the camera frame accurately and rapidly. A pressurized spray mechanism capable of accurately spraying a target a 1-meter range was developed, and a controller designed to maximize accuracy was implemented on an Arduino microcontroller. The system was mounted to a quadcopter and tested extensively.

Performance metrics were developed to evaluate the efficacy of test sprays. Testing was carried out to determine the effect of spray duration, range, UAV motion, and wind on these metrics. Overall, the system proved to be effective and robust. Representative targets and real weeds were sprayed under controlled conditions, yielding an overall spray accuracy of greater than 90%. Spraying was successful across a variety of ranges and spray durations, and proved unaffected by light wind conditions.

Skills:

• Project formulation and solution

• UAV design and operation

• Motion capture

• Computer vision

• Report writing and presentation

Takeaways:

My first full academic research project and publication was a formative experience – from developing a research question and translating it to actionable objectives, to the iterative R&D, and finally testing and analysis of the working system. The research-driven approach emphasized the importance of structured and systematic validation and testing – lessons I still apply to any engineering problem.