Submitted to: Transactions of the ASAE
Publication Type: Peer reviewed journal
Publication Acceptance Date: 12/9/2002
Publication Date: 6/1/2003
Citation: DELWICHE, S.R. CLASSIFICATION OF SCAB- AND OTHER MOLD-DAMAGED WHEAT KERNELS BY NEAR-INFRARED REFLECTANCE SPECTROSCOPY. TRANSACTIONS OF THE AMERICAN SOCIETY OF AGRICULTURAL ENGINEERS. 2003. Interpretive Summary: Fusarium head blight, also known as scab, is a fungal disease that occurs in small grains. In the United States it is most problematic in wheat (hard red spring, durum, and soft red winter classes) and barley. Scab may produce the mycotoxin deoxynivalenol (DON), also known as vomitoxin, which is toxic to non-ruminant animals. U.S. Food and Drug Administration advisory levels specify that DON in finished wheat products destined for human consumption should not exceed 1 part per million. Traditionally, official inspection for scab entails human visual analysis of sample of hundreds of kernels, thus requiring 10 minutes or longer per sample. In the current study, a near-infrared (NIR) spectrometer was used to examine the reflectance trace of intact wheat kernels, with and without scab damage. Additionally, a third category was examined, this being mold damage. A total of 577 kernels were examined by NIR reflectance, having first been categorized (sound, mold, or scab) by inspectors of the USDA Grain Inspection, Packers and Stockyards Administration (GIPSA). Classification trials, based on linear discriminant analysis, were developed by exhaustive searches of combinations of two to four NIR readings (using a total of 118 initial readings) that would produce the most accurate classification. The results indicated that effective classification (correctness in classification at 95% or higher) could be obtained by use of just two NIR readings. Still better results were obtained by including kernel weight into the classification models or by combining the mold and scab categories into one, thus producing a model that identifies kernels as either sound or damaged. The achieved accuracy levels demonstrate the feasibility of using NIR reflectance spectroscopy to assist in wheat grading and commercial sorting.
Technical Abstract: Scab (Fusarium head blight) is a disease that causes wheat kernels to be shriveled, underweight, and difficult to mill. Scab is also a health concern because of the possible concomitant production of the mycotoxin deoxynivalenol. Current official inspection procedures entail manual human inspection. A study was undertaken to explore the possibility of detecting scab-damaged wheat kernels by a near-infrared (NIR) diode array spectrometer. Wheat kernels from three categories-sound, scab-damaged, and mold-damaged-were visually inspected by, and furnished by the USDA Grain Inspection, Packers and Stockyards Administration (GIPSA). The reflectance spectrum of each intact kernel was collected at 6-nm increments over a 940-1700 nm region. Exhaustive searches of the best combination of individual wavelengths, best difference of wavelengths, best ratio, and combinations thereof were performed on a set of 100 kernels from each of the three categories. The best modeling classification occurred for precisely aligned kernels using a combination of kernel mass and the difference in log(1/R) at two wavelengths, 1182 and 1242 nm. When applied to a test set the classification model correctly identified the intended category with 95% accuracy. The most difficult to classify conditions were scab-damaged vs. mold-damaged and mold-damaged vs. sound. When scab-damaged and mold-damaged kernels were combined into one category, the overall accuracy for a two-category (sound vs. damaged) classification model was between 95% and 98%, depending on the kernel orientation scheme and inclusion of kernel mass. The achieved accuracy levels demonstrate the feasibility of using NIR reflectance spectroscopy with as few as two wavelengths to assist in wheat grading and commercial sorting.