Comparative study on detection of moisture content in chestnuts with shell using visible/near infrared,and mid-and short-wave near infrared spectroscopy
In order to detect and evaluate the moisture content of chestnuts with shells non-destructively and rapidly, the quantitative prediction methods based on visible/near-infrared, and mid- and short-wave near-infrared spectroscopy techniques were explored and compared. First, different preprocessings were applied to full spectra in wavebands of I and Ⅱ to establish PLSR models. Results show that SNV preprocessed spectra in waveband I and normalization preprocessed spectra in waveband Ⅱ were individually optimized. The Rp and RMSEP in prediction set achieved 0.882 and 0.856, and 1.389％ and 1.665％, respectively. In further steps, iPLSR, BiPLSR and SiPLSR modeling methods based on full spectra with the optimal preprocessing methods were also employed and compared. The models' performance indicates that waveband I is generally superior to waveband Ⅱ, and the iPLSR model performed best with Rp of 0.910 and RMSEP of 1.180％. In this study, the optimal waveband, preprocessing method, and modeling method for detecting the moisture content of chestnuts with shells using near-infrared spectroscopy were confirmed. These results provide important technical and methodological support for high-throughput online rapid detection of chestnut moisture content in the future.
chestnutmoisture contentpartial least squareswavebandnear-infrared spectroscopy
万方数据
维普
