基于改进卷积神经网络的苹果叶片病害识别
Apple leaf disease identification based on improved convolutional neural network
童子皓 1赵宏伟 1李蛟 2王紫薇3
作者信息
- 1. 吉林大学计算机科学与技术学院,长春 130012
- 2. 吉林大学图书馆,长春 130012
- 3. 吉林省商务信息中心,长春 130061
- 折叠
摘要
针对传统的卷积神经网络对农业领域的识别精度不高的问题,对现有的VGG16网络模型进行改进,用于解决苹果树叶片病害的识别与预测问题.网络通过对输入图像进行卷积操作获得特征图,首先在每一个卷积层结束后加入批量归一化模块,用于提高模型的收敛速度.同时将卷积层的ReLU激活函数替换为P-ReLU函数以便提高网络训练效率,解决梯度消失的问题.在公开的Pla-ntVillage数据集中的实验结果为98.578％,表明文中的改进方法较一些经典网络模型有了更好的病害分类与预测精度.
Abstract
To solve the problem that the recognition accuracy of the traditional convolutional neural network in the agricultural field is not high,the existing VGG16 network model is improved to solve the problem of the identification and prediction of apple leaf diseases.The network obtains the feature map through convo-lution operation towards the input image,and adds mass normalization module after each convolutional layer is completed to improve the convergence speed of the model.At the same time,the ReLU activation function of the convolutional layer is replaced with the P-ReLU function to improve the efficiency of network training and solve the problem of gradient disappearance.The experiment result in the published plantvillage dataset is 98.578％,which indicates that the improved method in this paper has better disease classification and prediction accuracy than some classical network models.
关键词
计算机应用/深度学习/卷积神经网络/病害识别Key words
computer application/Deep Learning/Convolutional Neural Network/disease identification引用本文复制引用
基金项目
吉林省省级科技创新专项资金项目(20190302026GX)
吉林省自然科学基金(20200201037JC)
出版年
2024
NETL
NSTL
万方数据