Scholarlink

In order to solve the problem of high labor cost and low grading efficiency caused by manual and mechanical apple grading, this paper compares ordinary image processing and three Apple system grading methods based on MobileNet V2 and ResNet 18 models. Firstly, the common image processing method was used to extract the features of apple size, shape and color, and grade the apples. After that, two apple classification models of MobileNet V2 and ResNet18 were built on window11, and the optimal model was obtained by self-built dataset and modification of the number of threads, learning rate, iteration and other parameters of the two models. The results of the comparison experiment between the two models and the Apple grading method of ordinary image processing show that the accuracy of the ResNet18 model is 98%, that of the MobileNet V2 model is 94%, and that of the ordinary image processing method is about 85%. The results show that the accuracy and efficiency of apple classification using ResNet 18 model are optimal.

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

Research on Apple System Grading Method Based on Deep Learning Abstract

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Zhiyu Li,
  • Qingxue Liu,
  • Jirong Zhang,
  • Meng Zheng,
  • Hongyu Xu,
  • Bo Wang,
  • Yun Su,
  • Fuyu Shen

摘要

In order to solve the problem of high labor cost and low grading efficiency caused by manual and mechanical apple grading, this paper compares ordinary image processing and three Apple system grading methods based on MobileNet V2 and ResNet 18 models. Firstly, the common image processing method was used to extract the features of apple size, shape and color, and grade the apples. After that, two apple classification models of MobileNet V2 and ResNet18 were built on window11, and the optimal model was obtained by self-built dataset and modification of the number of threads, learning rate, iteration and other parameters of the two models. The results of the comparison experiment between the two models and the Apple grading method of ordinary image processing show that the accuracy of the ResNet18 model is 98%, that of the MobileNet V2 model is 94%, and that of the ordinary image processing method is about 85%. The results show that the accuracy and efficiency of apple classification using ResNet 18 model are optimal.