Cherry tomato, which has been a must-have product in both summer and winter, is in competition with other crops with also its price. Therefore, measuring and predicting tomato quality non-destructively has been gaining interest together with the developments in automated classification systems. Non-destructive sensing systems applied one by one on each agricultural product have been providing enormous advantages by providing a whole quality prediction including inner and outer properties of the product. In this study, cherry tomato fruit color has been aimed to be predicted using FT-NIR spectroscopy measurements. For this purpose, it was aimed to predict fruit color properties of cherry tomatoes using FT-NIR spectroscopy in this study. First of all, FT-NIR spectrums of cherry tomatoes, which were grown in a green house and harvested at 6 different maturity stages, were taken using a FT-NIR spectrometer (Bruker Optik, GmbH, Ettlingen, Germany). After spectrum measurements, colors of the tomatoes were measured in L*a*b* color space using Minolta CR-400 (Konica Minolta Sensing, Osaka, Japan) equipment. In addition to L*a*b* color values measured, chroma and hue were also calculated using a* and b* values. Later on, prediction models were developed between measured FT-NIR spectrums and color values (L*, a*, b*, Chroma and Hue) using partial least squares (PLS) method and OPUS (Bruker Optik, GmbH, Ettlingen, Germany) software. In the result, color values were able to be predicted with high coefficients of determination using FT-NIR spectroscopic readings. A high coefficient of determination in predicting L* color value was obtained; it was 0.90 (RMSEP= 2.41) for validation data group while it was 0.92 (RMSEC= 2.17) for calibration data group. After the L* color value, the highest coefficient of determination was obtained for predicting a* color value; coefficient of determination obtained in predicting a* color value was 0.81 (RMSEP=4.21) for validation data group while it was 0.83 (RMSEC=4.02) for calibration data group. On the other hand, chroma and hue, which were calculated using a* ve b* color values, were also predicted with high coefficients of determination; in this stage, while Hue color value was predicted with a coefficient of determination of 0.72 (RMSEP= 0.31) in validation, it was predicted with R2=0.81 (RMSEC= 0.25) in calibration stage. Chroma, on the other hand, was predicted with R2=0.71 (RMSEP= 2.47) in validation while it was predicted with R2=0.81 (RMSEC= 2.19) in calibration. Although prediction results for b* color value was the lowest (R2=0.64, RMSEP= 3.33 for validation and, R2=0.79, RMSEC= 2.58 for calibration), it can still be assumed as a good model as the good PLS prediction results for agricultural applications can be as low as the results obtained for this application. As a result, successful prediction models were obtained in predicting color values of cherry tomatoes using FT-NIR reflectance spectroscopy measurements.

Anahtar Kelimeler: FT-NIR Reflectance Spectroscopy, PLS Analysis, Cherry Tomato, L*a*b* Color Values