Real-time Moisture Measurement on a Forage Harvester using Near Infrared Reflectance Spectroscopy

摘要

A mobile, diode array NIR spectrometer was integrated into the spout of a self-propelled forage harvester to measure crop moisture. Spectra and moisture reference samples were collected in 2004 and 2005 for the development of static (in-lab) and dynamic (on-harvester) moisture calibrations. Moisture prediction models for whole-plant-corn-silage (WPCS) developed using static data had a root mean standard error of cross validation (RMSECV) of 1.12% using five principle components (PCs) while a calibration developed using dynamic data had a RMSECV of 3.28% using four PCs. Alfalfa validation results produced RMSECVs of 2.50% using four PCs and 3.74% using three PCs for models using static and dynamic data, respectively. Dynamic data were predicted withcalibrations developed using static data with similar error but more spectral information was required. A static alfalfa model predicted dynamic data with a root mean standard error of prediction (RMSEP) of 3.41% using three PCs compared to the dynamic model's RMSECV of 3.74% using three PCs. Similar trends were found with WPCS models. Predicting data independent of type of crop resulted in the utilization of more PCs but with higher RMSEPs than the cross validation results of predicted dataset. The sensor and associated calibrations were able to predict forage moisture well, although more diverse data and further calibration development is needed to improve sensor accuracy to the desired range of +- 2 percentage units.