In-line estimation of falling number using near-infrared diffuse reflectance spectroscopy on a combine harvester

Abstract

Quality is an essential attribute of agricultural products and production processes. Wheat (Triticum aestivum L.) quality is primarily classified according to protein concentration and sub-classified depending on additional parameters, such as moisture content, sedimentation value and Hagberg falling number (HFN). Real-time sensing of grain protein concentration by means of near-infrared reflectance spectroscopy (NIRS) is an established method of assessing cereal grain quality during harvest. The objective of this study was to obtain NIRS calibration models for determining α-amylase activity of wheat and to identify changes of wheat quality. Performance characteristics were obtained during field trials in 2011 and 2012. HFN predictions correlated with reference measurements (R2 = 0.70). The standard deviation of differences between the NIR-predicted and reference values denoted as standard error of prediction was 37 s. Processed data were classified using principal component analysis, the prediction range of HFN and Hotelling T2-statistics. The average difference of NIR HFN estimation and HFN laboratory analysis was 34 s. The results obtained indicated that the use of near-infrared reflectance inline spectroscopy on combine harvesters can provide information for grain growers to optimize grain processing and marketing.