The Guelph Plot Analyzer: Semi-Automatic Extraction of Small-Plot Research Data from Aerial Imagery
1D. Drover, 2J. Nederend, 1B. Reiche, 2B. Deen, 2L. Lee, 1G. W. Taylor
1. School of Engineering, University of Guelph, Guelph, Canada
2. Plant Agriculture, University of Guelph, Canada
Small-plot trials are the foundation of open-field agricultural research because they strike a balance between the control of an artificial environment and the realism of field-scale production. However, the size and scope of this research field is often limited by the ability to collect data, which is limited by access to labour. Remote sensing has long been investigated to allocate labour more efficiently, therefore enabling the rapid collection of data. Imagery collected by unmanned aerial vehicles (UAVs) are a significant development in remote sensing for agricultural research, and their potential for efficient workflows has generated interest from agricultural scientists. However, data analysis techniques have not matured at the same rate, and a knowledge gap exists between end users of the data and those who can manipulate, extract, and deliver it. This study was established to address the barrier to adoption of UAVs created by this knowledge gap. We created a tool that can semi-automatically extract plot-level statistics from UAV-acquired imagery. This tool simplifies tasks that were previously accomplished via a Geographic Information System (GIS) by incorporating these tools into a web-based application, the Guelph Plot Analyzer (GPA). Users can upload a GeoTiff raster file to the application, and are presented with the UAV-acquired map, as well as a variety of polygon drawing tools. Using a hierarchy of Trial to Replication to Plot, the user draws boundaries around each category, and the tool can then automatically populate a shapefile with polygons corresponding to the plots. Polygons can be buffered to remove border effects, and alleyways can be specified to correctly align rows. Once finalized, the user can export the overlay as a shapefile, as well as a spreadsheet containing image statistics, including the mean, median, range, and a histogram of pixel values. The plots are labelled according to the user’s specified naming convention, making the data easily transferrable to statistical analysis software, as well as seamless to integrate into existing studies. The plot extraction tool is an efficient means for non-remote sensing scientists to turn qualitative imagery into quantitative measures and will help modernize small-plot research as UAVs become more common.