Research on Red Fuji Apple Grading Method Based on Improved Decision Tree

• Published in: 2024 4th Asia-Pacific Conference on Communications Technology and Computer Science (ACCTCS) pp. 707 - 712
• Main Authors: Zhang, Haitao, Cui, Xiaoxiao, Yang, Junhao, Junyao, W., Amulikemu, Samir, Wang, Yingchao
• Format: Conference Proceeding
• Language: English
• Published: IEEE 24.02.2024
• Subjects: Accuracy; Computational modeling; Computer science; Decision tree; Distributed databases; Image color analysis; K-means clustering; Pruning operation; Sensitivity; Shape; Threshold segmentation
• DOI: 10.1109/ACCTCS61748.2024.00131

In view of the problem that the grading time in apple grading research cannot meet the actual demand, the team proposed an apple grading method based on improved decision tree. Firstly, K-means clustering and threshold segmentation were used to separate the apple background. Then, five features of apple were extracted in sequence, including color, shape, diameter, texture and defects. We built a decision tree model based on the five external features of apples, and improved it with pruning operations. The results showed that the accuracy of the grading method for Red Fuji apples based on the improved decision tree was 96.75 % , which was 1.5 percentage points higher than the decision tree algorithm. This method can provide further scientific basis and theoretical methods for the research on grading of Red Fuji apples.