Prediction of the fibre content of sugarcane stalk by direct scanning using visible-shortwave near infrared spectroscopy

• Published in: Vibrational spectroscopy Vol. 101; no. none; pp. 71 - 80
• Main Authors: Phuphaphud, Arthit, Saengprachatanarug, Khwantri, Posom, Jetsada, Maraphum, Kanvisit, Taira, Eizo
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01.03.2019; Elsevier BV
• Subjects: Breeding programmes; Destructive testing; Fibre content; Infrared spectroscopy; Mathematical models; Near infrared radiation; Predictions; Short wave radiation; Spectra; Spectrum analysis; Sugarcane; Sugarcane stalks; Visible-shortwave near infrared spectroscopy; Sugarcane stalks; Breeding programmes; Visible-shortwave near infrared spectroscopy; Fibre content
• ISSN: 0924-2031; 1873-3697
• DOI: 10.1016/j.vibspec.2019.02.005

Knowledge of the fibre content in sugarcane stalks is important for breeding programmes and non-destructive measurements for its applications. This study aimed to use portable visible-shortwave near infrared (Vis/SWNIR) with a wavelength range of 570–1031 nm to evaluate the fibre contents of cane stalks. Fibre models were established by partial least squares (PLS) regression using spectra obtained from interactance mode with an absorbance unit of log (1/R). The models were constructed using a sample set of three sample sections (i.e., bottom, middle, and top of the sugarcane stalk). The samples were scanned with different integration times (200, 300, and 400 ms), and the spectra were pre-treated using different pre-processing techniques. The fibre content models were established based on both combined sample sections (CSS model) and individual sample sections (ISS model) obtained from the combination of three sample sections and an individual sample section, respectively. The results showed that the models that were developed using raw spectra with integration time of 300 ms had the best performance. This model had coefficients of determination of the prediction set (r2) of 0.75, 0.81, 0.81 and 0.71 and root mean square errors of prediction (RMSEP) of 0.81, 0.63, 0.80 and 0.73%fibre for the CSS model and bottom section, middle section, and top section of the ISS model, respectively. These results indicated that the models could be used for screening. Moreover, it was observed that the bottom section model had the lowest RMSEP. The model can be used as a rapid protocol for predicting the fibre content of sugarcane stalks, making it a useful method for a breeder to screen the fibre content in the field when monitoring during breeding programmes.