Scholarlink

<p>In this study, a&#xa0;model was developed using camera images and convolutional neural networks (CNNs) to classify ‘Fuji’, ‘Red Delicious’ and ‘Golden Delicious’ apple varieties. The model was used to distinguish between intact and rotten fruits without differentiating between apple varieties. Classification based on visual data provides a&#xa0;noncontact and fast evaluation, reducing labor requirements and increasing operational efficiency. The results obtained in the study show that the CNN-based approach provides high accuracy in the classification of apples. The model’s recall/sensitivity rates for the “Intact” and “Rotten” classes were 81.43% and 80.00%, respectively. In contrast, the overall accuracy rate of the model was 80.71%. The false positive and false negative rates were 19.72% and 18.57%, respectively, for the Intact class and 18.84% and 20.00%, respectively, for the Rotten class. These results show that the model is successful for both classes and has a&#xa0;strong generalization ability. In conclusion, this study highlights the application potential of deep learning-based automatic apple classification systems in the agriculture and food industry and suggests that they are important tools in quality control processes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Automatic Apple Classification Using Camera Footage and a Convolutional Neural Network

  • 期刊论文

  • 发表日期：

    2026年5月13日

查看全文
  • İbrahim Hakkı Kadirhanoğullari

摘要

In this study, a model was developed using camera images and convolutional neural networks (CNNs) to classify ‘Fuji’, ‘Red Delicious’ and ‘Golden Delicious’ apple varieties. The model was used to distinguish between intact and rotten fruits without differentiating between apple varieties. Classification based on visual data provides a noncontact and fast evaluation, reducing labor requirements and increasing operational efficiency. The results obtained in the study show that the CNN-based approach provides high accuracy in the classification of apples. The model’s recall/sensitivity rates for the “Intact” and “Rotten” classes were 81.43% and 80.00%, respectively. In contrast, the overall accuracy rate of the model was 80.71%. The false positive and false negative rates were 19.72% and 18.57%, respectively, for the Intact class and 18.84% and 20.00%, respectively, for the Rotten class. These results show that the model is successful for both classes and has a strong generalization ability. In conclusion, this study highlights the application potential of deep learning-based automatic apple classification systems in the agriculture and food industry and suggests that they are important tools in quality control processes.