Apple Grading Based on One-Vs-Previous Ordinal Binary Decomposition SVM Model

To enhance the sales competitiveness of apples, an apple grading system based on ordinal binary decomposition SVM model is proposed. Taking Yantai Red Fuji apples as the research subject, spectral data and Soluble Solid Content of apple samples are collected. With the spectral data as the model input and the sample grade as the output, the grading model is designed based on SVM models. Transforming ordinal variables into multiple binary ones, the One-Vs-Previous encoding scheme is designed for the binary decomposition SVM model. Experimental results demonstrate an average grading accuracy of 92.06%. The model improves both the predictive accuracy and robustness of apple grading.