Grain grading using electronic tools based on near-infrared spectroscopy and machine vision has the potential to replace current visual inspection methods. A compact and robust lab-scale spectrograph system that can be customized for agricultural applications was designed and tested. The spectrograph can detect spectral signals with wavelengths ranging from approximately 433 nm to 1067 nm. Spectral data in the short near-infrared region (866--1067 nm) and in the visible region (433--750 nm) can be acquired simultaneously. The spectrograph was configured to work under reflectance mode with external illumination and a full spectral scan required 0.7 s. Data acquisition and user interface programs were developed under LabVIEW environment.; A 650 nm single-mode diode laser was used to calibrate the spectrograph wavelength equation. An argon lamp was used to validate calibrated wavelength equation. A maximum wavelength error of 1.95 nm was determined. An optical bandpass expressed in full-width-at-half-maximum (FWHM) of 5.7 nm in the 433--750 nm, and 5.5 nm in the 866--1067 nm region was estimated experimentally. Repeated measurement of reflectance spectra of a Polytetrafluoroethylene (PTFE) disk provided coefficients of variation (CV) values below 0.2% within 30 s sampling period and remained under 0.4% within 30 min sampling period. To test the applicability of the spectral signals colleted using the spectrograph, moisture content of ground wheat samples was determined in the 866 nm to 1067 nm region. For an independent validation set, root mean squared error of prediction (RMSEP) of 1.3% and coefficient of determination (R 2) around 0.88 were achieved.
展开▼
