Title: RAPID SINGLE-KERNEL NIR MEASUREMENT OF GRAIN AND OIL-SEED ATTRIBUTES Author
Submitted to: Applied Engineering in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: June 1, 2006
Publication Date: September 1, 2006
Repository URL: http://asae.frymulti.com/toc_journals.asp?volume=22&issue=5&conf=aeaj&orgconf=aeaj2006
Citation: Armstrong, P.R. 2006. Rapid single-kernel nir measurement of grain and oil-seed attributes. Applied Engineering in Agriculture. Vol. 22(5):767-772. Interpretive Summary: Seed development requires the evaluation of hundreds of seed lines over multiple years to produce only a handful of commercial varieties or hybrids each year. Single-kernel near infrared (SKNIR) measurement has been used to improve and accelerate this process by being able to measure and sort seed for desirable characteristics. Current measurement rates of SKNIR systems are around 1 kernel/s which limits the amount of seed that can be physically processed. A new system was designed and tested that can measure seed characteristics for corn and soybeans at 10 kernels/s. Results show the instrument worked reasonably well for measuring corn attributes but even better for soybeans. The new SKNIR system has excellent potential for reducing the time and costs associated with the development of corn hybrids and soybean lines with specific composition or processing traits. Future work will focus on developing a mechanical feeding and sorting mechanism and methods to further improve measurement accuracy.
Technical Abstract: A single-kernel near infrared (SKNIR) instrument was designed and tested for fast measurement of corn and soybean attributes. The design was centered on achieving a spectral collection rate of 10 kernels per second which, limited integration times of the spectrometer to 30 ms. A spectrum of an individual kernel was collected as it slid along the length of a glass tube and was illuminated by multiple lamps. PLS regression models, developed to predict constituents from spectra resulted in models with an SECV of 0.93% MCdb for corn moisture, 0.33% MCdb for soybean moisture, and 0.99% for soybean protein. RPD values ranged from 3.1 to 5.5 for corn moisture and from 2.3 to 4.9 for soybean protein. Soybean moisture prediction was more quantitative with RPD values ranging from 7.2 to 10.5. Multiplicative scatter correction improved predictions for soybean moisture and protein but not for corn moisture. These results indicate that reasonable predictions can be made at fast NIR scan rates.