» Articles | Agricultural Journals

This study aimed at determining the relationships between the selected quality properties of four new apple clones: ‘Free Redstar’ × ‘Melfree’, ‘Free Redstar’ × ‘Ariwa’, ‘Free Redstar’ × ‘Rajka’, and ‘Free Redstar’ × ‘Rubinola’. Significant differences between the selected clones in terms of quality properties were revealed. For example, the ‘Free Redstar’ × ‘Ariwa’ clone had the highest total sugar content (132.7 g kg⁻¹) and total polyphenol content (TPC) (260.7 mg 100 g⁻¹ GAE), while ‘Free Redstar’ × ‘Rajka’ was characterized by the highest total acid level of 739.0 mg 100 g⁻¹ and the lowest a* value of −0.62, which were significantly different from all other clones. Correlations between image textures and chemical and color characteristics were observed. Among the sugars, sucrose showed the strongest positive correlation with the selected texture parameters of images in color channel Y, obtaining a Pearson correlation coefficient (R) of 0.946, while sorbitol had the strongest negative correlation (R = −0.946) with the selected texture of images in color channel b. In terms of acids, the Pearson correlation coefficient was up to 0.953 for the relationship between the content of citric acid and the selected texture of images in color channel G, and for TPC, the value of Pearson’s correlation coefficient reached 0.921 for the selected texture feature of images in color channel X. Regarding color parameters, the correlations were also strong, reaching 0.945 between parameter a* and the selected texture of the images in color channel a. These findings provide insights into the potential use of image texture analysis in combination with chemical profiling for evaluating apple quality and clone differentiation.

The objective of this study was to evaluate the usefulness of image analysis and traditional machine learning algorithms for the discrimination of raspberry cultivars originating from organic and conventional cultivation systems. Images of the whole fruit of four raspberry cultivars: Polana, Polesie, Poemat, and Poranek, were acquired using a digital camera. Texture descriptors were extracted from images. The mean values of selected image features were compared, and the artificial intelligence models were developed. The mean comparison revealed significant differences in image texture descriptors among raspberry cultivars within both cultivation systems and between organic and conventional samples of the same cultivar. For cultivar discrimination of Polana, Polesie, Poemat, and Poranek in organically grown raspberries, the highest accuracy of 89.2% was achieved using the Subspace Discriminant classifier, followed by Quadratic Support Vector Machine (88.1%) and Wide Neural Network (86.0%). In conventionally cultivated raspberries, the highest accuracy of distinguishing cultivars reached 91.7% for the Linear Discriminant model, followed by Subspace Discriminant (91.3%) and Medium Neural Network (88.9%). All models were characterized by high area under the Receiver Operating Characteristic (ROC) curve (AUC) values, indicating excellent discrimination performance. Complete separation of organic and conventional raspberries for all analyzed cultivars with an accuracy of 100% was obtained. The results demonstrate that image texture analysis combined with traditional machine learning algorithms provides a rapid, objective, and non-destructive approach for raspberry cultivar identification and cultivation-system authentication. The proposed methodology shows considerable potential for applications in quality control, traceability, and verification of production practices in the raspberry industry.