Submitted to: Sensing and Instrumentation for Food Quality and Safety
Publication Type: Peer reviewed journal
Publication Acceptance Date: 1/21/2008
Publication Date: 6/12/2008
Citation: Delwiche, S.R. 2008. High-speed bichromatic inspection of wheat kernels for mold and color class using high-power pulsed leds. Sensing and Instrumentation for Food Quality and Safety. 2(2):103-110. Interpretive Summary: In the United States, the authority to regulate mycotoxins, inclusive of deoxynivalenol (DON), a by-product of the fungal disease Fusarium Head Blight (FHB), is codified in the Federal Food, Drug and Cosmetic Act, which places authority with the Food and Drug Administration (FDA). Certain mycotoxins, such as aflatoxin, a recognized carcinogen, are regulated through action levels, which can then necessitate official testing for the mycotoxin and can result in the condemnation of grain lots in excess of the action level. Other mycotoxins, including DON, are not regulated by FDA, per se, but instead are voluntarily controlled under the guidelines of advisory levels. Depending on the intended use, the advisory level for DON in the United States ranges from 1 mg/kg (finished product for human consumption) to as much as 10 mg/kg (in certain animal feeds). The current study describes the use of three forms of optical measurement of single wheat kernels for screening of Fusarium head blight for eventual incorporation in high-speed optical sorters. Previous research demonstrated a sorting efficiency of approximately 50 percent with existing high-speed equipment, but a much higher efficiency (~95%) when analytical spectrometers are used. The intention of the current work is to bridge this efficiency gap. Knowledge gained from analysis of the single kernel in-flight response will provide design criteria for improvement of high-speed optical sorters for recognition of mold-damaged wheat. With as few as two wavelengths, (e.g., 500 nm and 550 nm), stationary spectrometer-based classification systems are approximately 95% accurate in identifying normal and Fusarium-damaged kernels. Secondly, time-domain waveforms of the reflected energy from freefalling kernels, as sensed by a fiber optic probe, show differences between normal and Fusarium-damaged kernels on average; however, such waveforms are influenced by the random orientation of the kernel. Lastly, improvements in classification of freefalling kernels should arise with the combination of simultaneous capture of reflected light at two wavelengths and the collection of multiple reflectance readings per wavelength as the kernel traverses the detector field of view. Mills, processing plants, optical instrumentation manufacturers, and official grain inspection programs are the intended beneficiaries of this research.
Technical Abstract: High-speed optical sorting of seeds in commercial processing is routinely practiced for removal of discolored seeds, seeds from volunteer plants, and non-seed objects. Sorters are conventionally based on monochromatic or bichromatic light from broad wavebands in the visible and near-infrared regions of energy. A particular challenge for these devices has been the recognition and removal of wheat kernels that have been damaged by the mold caused by the fungal disease Fusarium Head Blight. Previous research using an off-the-shelf bichromatic design on Fusarium-damaged wheat kernels demonstrated that approximately half of damaged kernels were positively detected. The research described herein examines an alternative design for bichromatic lighting and applies this design to two scenarios: sound vs. Fusarium-damaged wheat and red vs. white wheat. The new design utilizes two high-brightness (HB) LEDs and one silicon photo diode detector. The LEDs are flashed in alternating sequence at high frequency (2000 Hz), such that during the half-cycle time period (0.25 ms) that each LED is on, reflected energy readings at a 10' sampling frequency are captured from a kernel in flight. This permits the capture of approximately 20 cycles of pulsed light during the time the free-falling kernel passes through the field of view of a fiber optic probe. A linear discriminant analysis (LDA) classification algorithm was developed that used two values derived from the reflected energy readings. Based on the new design, the accuracy of sound vs. Fusarium-damaged classification was 78% on average; for red vs. white wheat classification, the average accuracy was 76%. The slope parameter was the dominant classifier for both sets of wheat kernels. Although these accuracy values are not at the level as that obtained from LDA models that utilize reflected energy readings at two wavelengths from stationary kernels (95% and 92% for sound vs. Fusarium-damaged and red vs. white, respectively), the new design offers an improvement over conventional bichromatic designs.