Today we are excited to announce the release of new tools for analysis of field trials and custom field zones. Our new tools make it quick and efficient to extract boundaries of each individual plot in the trial area or define a custom zone and calculate a number of statistics for any vegetation index or raw sensor measurements.
Data collection and evaluation of agricultural field trials are labor intensive and time-consuming. By using drones equipped with RGB or multispectral sensors, one can now map trial areas in a matter of minutes and collect detailed spectral data over each plot in the field trial. While this approach significantly reduces time spent in the field, it still requires considerable efforts to digitize plot boundaries and make it possible to derive plot-wise statistics from orthomosaics.
Traditionally one would use GIS software like QGIS or ArcMap to create a template polygon and copy-paste it manually and place it so it fits boundaries of other plots. While this approach gives full control over the accuracy it may be very time consuming in large trials with hundreds or even thousands of plots. Our new tool automates this whole process. With minimal input from the user, the tool makes use of machine learning algorithms to automatically detect boundaries of each plot and calculates summary statistics for individual bands and selected vegetation indices:
Zonal statistics tool also allows to define completely custom zones of any shape which can be used in fields with non-regular trial layout or in virtually any other field where it might be of interest to compare the performance of specific parts:
Plot-wise statistics derived from the detected boundaries include such metrics as minimum, maximum, mean, median, standard deviation, 25th, and 75th percentile values as well as a total number of pixels used for calculations. This data along with boundary polygons can then be exported in a geographical format ESRI shapefile for further analysis in GIS software or as a CSV file:
After extensive tests with various crops at different growth stages, we are excited to add this new feature to our toolbox that helps users gain even more value and insights from their drone imagery.
The Swedish University of Agricultural Sciences(SLU) that we’ve been working closely with during the development of this tool generously provided example datasets from their trials. Have a look at it at this link or try it with your own imagery and let us know what you think! This guide shows in details how it works.
The trial plot extraction tool has been developed within the SLU programme Laboratory for Intelligent Decision Support Systems (LADS) and with financial support from Vinnova(dnr: 2016–04248).