ON–THE–GO SENSING OF CHLOROPHYLL STATUS IN CORN

Abstract

An on–the–go system for sensing chlorophyll status in corn using neural networks and fiber–optic spectrometry, which was used to acquire spectral response patterns (SRPs), was developed and tested at a speed of 0.6 km/h to acquire data on five plots of corn. The speed of 0.6 km/h was used because of the low download speed of SRPs from the spectrometer to the computer and the need to obtain a large number of SRPs per unit distance. A neural network model incorporated into the mobile system was trained using statically collected plant–center SRPs and chlorophyll readings acquired by a SPAD 502 chlorophyll meter on the same day and in the same field plots. The model showed good correlation between predicted SPAD chlorophyll readings based on statically acquired SRPs and actual chlorophyll readings (r2 = 0.85, RMSE = 1.82 SPAD units). The developed neural network model for predicting chlorophyll readings was used to predict chlorophyll readings using dynamically acquired SRPs at 0.6 km/h. The RMSE values between plot average chlorophyll and predicted chlorophyll readings of mobile SRPs were less than 1.04 SPAD units, which were less than the RMSE value of the neural network model (1.82 SPAD units).