Non-destructive evaluation by Vis-NIR spectroscopy and PLS regression of soluble solids in several stone fruits..

摘要

In known applications of Vis-NIR spectroscopy in non-destructive postharvest quality evaluation of fruits either a full spectrum PLS regression or a few-wavelength MLR are adopted. However, dealing with very complex matrices such as intact fruitsaccuracy at the 0.5-0.6鳥rix level over a range of values at least 5-6鳥rix wide are difficult to obtain. In this contribution experimental results are presented demonstrating the advantage of wavelength selection practices in the development of PLS regression-based calibrations for SSC of cherries, apricots, and peaches. As regards cherry, a simple spectral selection scheme (one every five lines acquired) led to satisfactory results considering only the Vis-Herschel wavelengths from 600 nm to 1100 nm.The SEP was 0.49鳥rix against a SSC standard deviation of 2.96鳥rix and a range of 15.13鳥rix. Apricot represented a more interesting case, with a SEP improvement from 0.66鳥rix to 0.43鳥rix over full spectrum PLS using exactly the same pretreatments (first derivative, 6 point gap and 3 point segment, Weighted MSC) and 9 PLS factors. In addition, retaining 7 distinct segments, only a Vis-Herschel calibration can be determined. The whole data set (162 apricots) had a SSC standard deviation of 1.95鳥rix and a range of 9.55鳥rix. The last fruit considered is peach: selecting a big number of segments, up to 19, a SEP of 0.40鳥rix was obtained for 9 PLS factors. The whole data set (108 peaches) had a SSC standard deviation of about 1.64鳥rix and a range of7.92鳥rix.