|Long, Daniel - Dan|
Submitted to: Columbia Basin Agricultural Research Annual Report
Publication Type: Experiment Station
Publication Acceptance Date: 5/18/2006
Publication Date: 6/1/2006
Citation: Long, D.S., Baker, A.A. 2006. On-combine sensing of grain protein concentration in soft white winter wheat. Dryland Agricultural Research Annual Report. Oregon Agric. Exp. Sta. Special Report 1068. Interpretive Summary: This study investigated a new optical sensor designed to operate on a combine harvester for mapping grain protein levels of wheat. Reference samples of soft white winter wheat, obtained by hand from the exit auger of a combine, were analyzed for protein concentration and statistically compared with sensor measurements obtained during harvest. Results show that on-combine sensing is sensitive to spatial variability in grain protein, especially when within-field variability in protein is more than twice the instrument error. The results are sufficiently promising to suggest that on-combine NIR sensing is a potentially useful tool for mapping purposes.
Technical Abstract: The objective of this study was to evaluate an on-combine optical sensor for measuring and mapping the grain protein concentration in soft white winter wheat. Protein concentration was measured in reference samples of grain collected by hand from the combine during harvest. These reference measurements were statistically compared with sensor measurements that coincided nearly in time with the hand samples. Good linearity was observed in a scatter plot of sensed protein vs. reference protein (R^2 = 0.71). The standard error of prediction was relatively large (0.90%), which indicated that extremes in protein concentration would likely need to be more than twice this error, or 1.8%, for the sensor to detect differences within farm fields. The large error may be due to the effects of vibration and dust on the combine, or time difference between NIR measurement and reference sampling was too large. Precision will be further validated as a result of our comprehensive data validation activities in 2006.