Development of an unmanned aerial vehicle and image processing techniques for agricultural applications

Abstract

In this study, an unmanned aerial vehicle (UAV) was built for use in small breeding plots. The UAV was constructed from readily available off-the-shelf components, such as the Cube Orange flight controller and HERE3 GPS. The UAV was a hexarotor design powered with a 4s 9000mAh battery. The UAV underwent a series of tests across multiple dates to determine flight path accuracy. Two flight heights (30.48m and 60.96m) were taken on multiple dates. A custom pan tilt servo gimbal was constructed capable of carrying multiple cameras in different configurations. The gimbal was tested using a series of different weights and mounting locations. A case study was completed to demonstrate usefulness of the system and the open-source image processing pipeline which can be used with images captured from the custom built UAVs. In testing, the average lateral error of the UAV was 0.2m. The average vertical errors were 0.06m, and 0.33m according to the data with the extended Kalman filter (EKF) and the built-in barometer, respectively. The average errors of the gimbal at maximum load was 1.2 degrees in the pitch direction and 1.7 degrees in the roll direction. The case study demonstrated that free, open-source software can be successfully used to process images from the UAV system.