摘要
为进一步提高高光谱图像分类的精度与准确度,文章提出了一种基于深度卷积神经网络的图像分类方法,构建了适合高光谱数据的网络结构,并通过循环学习率与小批量梯度下降法来实现分类算法的改进,从而来提高图像分类算法的计算效率.通过训练测试,实验结果显示,高光谱图像分类算法与传统的分类方法相比,有着更高的准确率,即使在训练样本较少的条件下也有着良好的分类性能.
In order to further improve the precision and accuracy of hyperspectral image classification,this paper proposes a convolution of the neural network based on depth image classification method,constructs a network structure suitable for hyperspectral data,and improves the classification algorithm by the vector loop and small batch gradient descent,so as to improve the computational efficiency of image classification algorithm.Through the training test,the experimental results show that the hyperspectral image classification algorithm has higher accuracy compared with the traditional classification method,and has good classification performance even under the condition of less training samples.
作者
郑庆翔
朱敏
ZHENG Qing-xiang;ZHU Min(Modern Educational Technology Centre,Meizhou bay vocational and technical college,Putian 351254,China)
出处
《白城师范学院学报》
2020年第2期24-29,共6页
Journal of Baicheng Normal University
基金
福建省教育厅2018年中青年教师教育科研项目(JZ18103).
关键词
深度卷积神经
高光图谱
网络结构
分类性能
deep convolutional nerve
highlight map
network structure
classification performance