| 李国强,王俊妍,王天雷.基于共有特征学习和数据增强的农作物病害识别算法[J].高技术通讯(中文),2023,33(3):261~270 |
| 基于共有特征学习和数据增强的农作物病害识别算法 |
| Crop disease recognition algorithm based on common feature learning and data augmentation |
| |
| DOI：10. 3772/ j. issn. 1002-0470. 2023. 03. 004 |
| 中文关键词: 农作物病害;深度残差网络;数据增强;注意力机制;共有特征学习 |
| 英文关键词: crop diseases, depth residual network, data augmentation, attention mechanism, common featurelearning |
| 基金项目: |
| 作者 | 单位 | | 李国强 | (燕山大学电气工程学院 秦皇岛071000) | | 王俊妍 | (燕山大学电气工程学院 秦皇岛071000) | | 王天雷 | (燕山大学电气工程学院 秦皇岛071000) |
|
| 摘要点击次数: 984 |
| 全文下载次数: 770 |
| 中文摘要: |
| 针对农作物病害图像类间差异较小,传统机器学习方法在农作物病害数据集上识别精度低、模型训练复杂等问题,本文提出一种基于共有特征学习和数据增强的农作物病害识别算法。首先,对于农作物病害数据集类间数据不均衡等问题,本文使用Mixup 数据增强算法对数据集进行扩充,丰富样本数量;然后,对于特征提取模块,本文在深度残差网络中嵌入通道注意力模块,使之侧重学习农作物叶片病害特征,忽略背景信息对模型带来的干扰;最后,在提取完图像特征后,将特征图送入到共有特征学习模块中,提高图像之间线性关联,增强模型泛化性能和鲁棒性。为验证所提模型的有效性和实用性,本文在Plant Diagnosis Dataset 农作物病害数据集上进行训练及测试,实验结果表明,所提模型准确率达到97. 9%,可有效提高农作物病害图像识别精度。 |
| 英文摘要: |
| To address the problems of small difference between crop disease images, low recognition accuracy and complex model training of traditional machine learning methods in crop disease dataset, this paper proposes a crop disease recognition algorithm based on common feature learning and data augmentation. Firstly, for the problem of unbalanced data between classes in crop disease dataset, this paper uses Mixup data augmentation algorithm to expand the dataset and enrich the number of samples; then, for the feature extraction module, this paper embeds the channel attention module in the deep residual network, so that it focuses on learning the crop leaf disease features and ignores the interference brought by the background information to the model. After extracting the image features,the feature maps are fed into the common feature learning module to improve the linear correlation between images and enhance the generalization performance and robustness of the model. In order to verify the effectiveness and practicality of the proposed model, the training and test are conducted on the Plant Diagnosis Dataset, and the experimental results show that the accuracy of the proposed model reaches 97. 9%, which can effectively improve the accuracy of crop disease image recognition. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
