| Popis: |
Weedy rice is one of the most well-known weeds in rice growing countries, particularly in Southeast Asia. Weedy rice, particularly paddy seed, is difficult to handle and separate because it has characteristics (morphological likeness) similar to those cultivated rice. This report describes a modification of self-organizing maps (SOMs) for categorizing weedy rice from cultivated rice using near-infrared spectroscopy (NIR) on paddy seed in situ. The sample pretreatment was performed using a cyclone vacuum machine to remove the contaminated particles and other impurities. Optical microscope and thermogravimetric analysis were employed to evaluate the physical properties and thermal behavior of rice sample, respectively. For direct sample analysis, a near-infrared spectroscopy with a reflectance attachment was used. To show the relevant NIR regions, the obtained NIR spectra were smoothed using the Savitzky–Golay polynomial, baseline-aligned using the standard normal variate, and mean-centered; in addition, the second derivative was determined. SOMs were well-optimized and applied for the classification of weedy samples from four cultivated rice. To generate a robust prediction, the predictive modeling was generated from 100 different training and test set split. The results achieved a very high predictive value in the range of 91–99% for precision and 88–99% for accuracy of the test samples, respectively. |
Full Text from ScienceDirect