A case study of a precision fertilizer application task generation for wheat based on classified hyperspectral data from UAV combined with farm history data

Different remote sensing methods for detecting variations in agricultural fields have been studied in last two decades. There are already existing systems for planning and applying e.g. nitrogen fertilizers to the cereal crop fields. However, there are disadvantages such as high costs, adaptability, reliability, resolution aspects and final products dissemination. With an unmanned aerial vehicle (UAV) based airborne methods, data collection can be performed cost-efficiently with desired spatial and temporal resolutions, below clouds and under diverse weather conditions. A new Fabry-Perot interferometer based hyperspectral imaging technology implemented in an UAV has been introduced. In this research, we studied the possibilities of exploiting classified raster maps from hyperspectral data to produce a work task for a precision fertilizer application. The UAV flight campaign was performed in a wheat test field in Finland in the summer of 2012. Based on the campaign, we have classified raster maps estimating the biomass and nitrogen contents at approximately stage 34 in the Zadoks scale. We combined the classified maps with farm history data such as previous yield maps. Then we generalized the combined results and transformed it to a vectorized zonal task map suitable for farm machinery. We present the selected weights for each dataset in the processing chain and the resultant variable rate application (VRA) task. The additional fertilization according to the generated task was shown to be beneficial for the amount of yield. However, our study is indicating that there are still many uncertainties within the process chain.