Multispectral Imaging And Elevation Mapping From An Unmanned Aerial System For Precision Agriculture Applications

Abstract: As the world population continues to grow, the need for efficient agricultural production becomes more pressing.  The majority of farmers still use manual techniques (e.g. visual inspection) to assess the status of their crops, which is tedious and subjective.  This paper examines an operational and analytical workflow to incorporate unmanned aerial systems (UAS) into the process of surveying and assessing crop health.  The proposed system has the potential to significantly reduce time, labor and cost while also yielding more accurate results, allowing farmers to better estimate their yield and obtain quantifiable data on troubled areas. The airframe for this study was built from a combination of hobby-grade and scientific components.  The aircraft incorporate avionics such as a Pixhawk autopilot system, GPS, and data telemetry links.  This allows for completely autonomous flight paths to obtain coverage.  The main sensor packages evaluated on the UAS for this study were a digital camera and a multi-spectral imager. Overlapping photos were taken during flight to ensure that there were no gaps in data.  Post flight the pictures were geo-located in a world fixed frame (e.g. WGS-84).  Data was collected across several flight tests conducted in Brisbane, Australia at the Samford Ecological Research Facility (SERF).  The primary output was a georeferenced, orthomosaic of the area in the visible light spectrum. Corresponding normalized difference vegetation index (NDVI) maps to assess vegetation health and vigor as well as a digital elevation models (DEM) that represents the terrain’s surface in 3D were developed as well.  The paper describes the process of obtaining and analyzing these results and compares data products generated using software such as QGIS, MicaSense Atlas, and Agisoft Photoscan Professional.  Additionally, analysis accuracy, best practices, and improvements of this type of aerial surveying are discussed.

read the paper

Authors: C. Lum, M. Mackenzie, C. Shaw-Feather, E. Luker, M. Dunbabin

Associations: University of Washington, Queensland University of Technology