This paper presents the validation of a digital image analysis method for the quantitative determination of physical quality parameters of apples by comparative analysis with a traditional measurement method. The method was used to determine the quality indicators of apples based on digital image analysis in accordance with standard requirements. Five popular apple varieties from Kazakhstan were selected for the study: Aport Alexander, Ainur, Sinap Almaty, Nursat and Kazakhskij Yubilejnyj. The parameters of the five apple varieties were measured both manually and digitally, which revealed close agreement between the obtained values. The analysis of the results of measuring the geometric parameters of the apples and the percentage of red color in the images was carried out. The maximum relative errors in determining the diameters (d, D) and height (h) were 2.99%, 3.03% and 4.12%, respectively. Regression models were developed to predict the weight and volume of apples. The best results in weight prediction were obtained for the Sinap Almatinsky variety using stepwise linear regression (R² = 0.96), and volume prediction showed the best results for the Nursat variety (R² = 0.92). This study lays the foundation for the development of automated systems for sorting apples by commercial varieties.

Application of pre-trained CNN and transfer learning approach for identification of different varieties of Kazakhstan apples is presented in the paper. The study used the most popular Kazakhstan apple varieties as: Aport Alexander, Ainur, Sinap Almaty, Nursat and Kazakhskij Yubilejnyj. In this study, two pre-trained CNN networks, SqueezeNet and GoogLeNet, were fine-tuned using different values of the Initial Learning Rate (ILR) parameter of 0.0001, 0.0002, 0.00025, 0.0003, 0.00035, 0.0004, 0.0005 and three network tuning algorithms, respectively The Stochastic Gradient Descent with moment solver (Sgdm), the Adam optimization algorithm (Adam), and Root Mean Square Propagation (RMSprop). The performance of the networks was evaluated by analyzing the values of Training Accuracy (TA), Training Loss (TL), Validation Accuracy (VA), Validation Loss (VL), and Confusion Matrix. The training of the networks was performed in 1900 iterations and 30 epochs. The developed approach for the identification of five varieties of Kazakhstan apples using deep learning techniques achieved 100% correct classification of fruits for the Ainur variety with a GoogLeNet network, solver RMSprop and ILR= 0.0005. For the varieties Aport, Kazakhski Yubileinyi and Nursat, one of the three network evaluation indicators achieves 100% accuracy, and for Sinap Almatynski – all three indicators are with and above the value of 95%. For varieties Ainur, Aport, Kazakhski Yubileinyi and Sinap Almatynski as the best (with the highest optimal accuracy values) is the GoogLeNet network, with the following settings: solver RMSprop and ILR=0.0005. Only for variety Nursat is the SqeezeNet network suitable, with the following settings: solver Sgdm and ILR= 0.0003. The proposed approach for recognizing the varietal affiliation of apples using deep learning neural networks is suitable for the analyzed apple varieties and could be easily implemented and used in industrial conditions for sorting fruits. The achieved recognition accuracy meets the requirements in the field.

This paper presents the validation of a numerical method for quantifying the physical quality parameters of apples through a comparative analysis with a traditional measurement method. The numerical method was applied to determine the parameters of batches of Kazakh apples according to standard requirements, using image analysis of apples. Five common varieties of Kazakh apples were selected: Aport Alexander, Ainur, Sinap Almatynski, Nursat and Kazakhski Yubileinyi. The geometric parameters of the apples and the percentage of red in the images were determined. The parameters of the 5 apple varieties were processed and measured both manually and digitally, revealing a close agreement between the obtained values. The developed digital method achieved high accuracy in determining the (diameters (d) and (D) in two perpendicular planes and height (h) of each apple), with maximum relative errors of 2.99% for (d), 3.03%, and 4.12% for the (h), and (D) parameters, respectively. Regression models were developed to determine and predict the mass and volume of apples via the digital method. The best results for the apple weight prediction were obtained for Sinap Almatynski variety by stepwise linear regression, and for the apple volume prediction were obtained for Nursat variety by linear regression. Regression equations for mass, volume and geometric dimensions constitute the basis for the development of a small instrument for automatically sorting apples by commercial variety.