Crop maturity mapping using a low-cost low-altitude remote sensing system

Abstract

The objective of this study was to assess the ability of the 'low-cost low-altitude (LCLA) remote sensing system' to map the maturity of a barley crop. Monitoring maturity is important from a frost/pest/disease susceptibility perspective. It also allows harvest to be planned, and in this case, screens varieties for adaptation to potentially tough seasons. The study area, a barley variety trial, was at 'Lundavra' near Goondiwindi in Southern Queensland (-28.056º, 150.087º). The LCLA remote sensing system consisted of digital cameras which, along with controlling electronics, were positioned in an unmanned aerial vehicle (UAV). The range of growth stages present varied from Zadok 43–59. Areas-of-interest were randomly selected from the variety plots, and a statistical package utilised to perform discriminant function analysis of the spectral values. The classification results (when predicting the original 14 classes) indicated that the predictive power was weak, with 23% correctly classified. As each class represents an individual growth stage of the crop, a difference of one in the Zadok scale can mean as little as an extra leaf unfolded on the plant. The accuracy was further improved by broadening the groupings to six secondary growth stages, three principal growth stages, and finally refining the classification to the two primary growth stages i.e. booting (Z40–49) and emergence (Z50–59). This resulted in a classification accuracy of 83.5%. The classification results achieved with the LCLA remote sensing system was quite acceptable, especially considering that the image was taken over a month after the growth stages were recorded.