您的位置：首页 > 论文页面
基于迁移学习的果蔬质量分类研究
发表时间：2021-06-19 浏览量：891 下载量：136
| 全部作者： | 郑凯，方春 |
| 作者单位： | 山东理工大学计算机科学与技术学院 |
| 摘 要： | 针对传统果蔬质量分类模型特征提取困难、训练耗时长、分类准确率低等问题，提出一种基于支持向量机（support vector machine，SVM）和预训练的VGG16深度学习网络混合模型图像分类方法。即在迁移学习的基础上使用SVM代替VGG16网络的全连接层及Softmax层进行分类，形成新的特征提取-特征分类（VGG16-SVM）模型结构。运用此方法在测试集上的分类准确率为99.1%，AUC（area under ROC curve）值达到0.999 6. 通过与普通卷积神经网络及各机器学习模型比较，验证了此方法能有效提高果蔬质量分类模型的训练速度与分类准确率。 |
| 关 键 词： | 计算机应用；VGG16深度学习网络；迁移学习；支持向量机；图像分类 |
| Title： | Research on quality classification of fruits and vegetables based on transfer learning |
| Author： | ZHENG Kai, FANG Chun |
| Organization： | School of Computer Science and Technology, Shandong University of Technology |
| Abstract： | Aiming at the difficulties in feature extraction, long training time and low classification accuracy of traditional fruits and vegetables quality classification models, an image classification method based on support vector machine (SVM) and pre-trained VGG16 deep learning network hybrid model is proposed. On the basis of transfer learning, SVM is used to replace the fully connected layer and Softmax layer of the VGG16 network for classification, forming a new feature extraction-feature classification (VGG16-SVM) model structure. Using this method, the classification accuracy on the test set is 99.1%, and the AUC (area under ROC curve) value reaches 0.999 6. Compared with ordinary convolutional neural networks and various machine learning models, it is verified that this method can effectively improve the training speed and classification accuracy of fruits and vegetables quality classification models. |
| Key words： | computer applications; VGG16 deep learning network; transfer learning; support vector machine; image classification |
| 发表期数： | 2021年6月第2期 |
| 引用格式： | 郑凯，方春. 基于迁移学习的果蔬质量分类研究[J]. 中国科技论文在线精品论文，2021，14（2）：220-228. |
1
评论数 0请您登录
暂无评论
下载全文
多维论文