Classification of the pits of different sour cherry cultivars based on the surface textural features

Abstract

The aim of this study was to develop the discriminative models based on selected textures of the outer surface of pit images to distinguish the different cultivars of sour cherry. The application of the image analysis technique ensured non-destructive, inexpensive and objective research. The textures for the images in the color channels R, G, B, L, a, b, X, Y, Z of the pits of four sour cherry cultivars ‘Debreceni botermo’, ‘Kelleris’, ‘Łutówka’ and ‘Nefris’ were calculated. In the case of discrimination of pits of all four cultivars, the accuracy was up to 96.25%. Slightly lower accuracies were observed for models built based on textures selected from Lab color space (94.12%) and color channel L (83.62%). However, all models allowed to completely distinguish the pits ‘Łutówka’ (100% correctly classified cases) from pits of other cultivars. In the case of analysis performed for pairs of cultivars, fully discrimination of the pits ‘Łutówka’ against the other pits was confirmed and the accuracy of 100% was determined for pairs of ‘Łutówka’ and ‘Debreceni botermo’, ‘Łutówka’ and ‘Kelleris’, ‘Łutówka’ and ‘Nefris’. The pits ‘Debreceni botermo’ and ‘Nefris’ were distinguished with the accuracy of up to 99% for the discriminative models built based on a set of textures selected from all color channels (R, G, B, L, a, b, X, Y, Z) and based on a set of textures selected from Lab color space. The accuracy reaching 98% was observed for distinguishing the pits ‘Kelleris’ and ‘Nefris’, and 95% for the pits ‘Debreceni botermo’ vs. ‘Kelleris’, in the case of models including textures selected from all color channels (R, G, B, L, a, b, X, Y, Z).