光学学报, 2018, 38 (11): 1128001, 网络出版: 2019-05-09   

基于集成卷积神经网络的遥感影像场景分类 下载: 1276次

Scene Classification of Remote Sensing Images Based on Integrated Convolutional Neural Networks
作者单位
1 中国科学院长春光学精密机械与物理研究所, 吉林 长春 130033
2 中国科学院大学, 北京 100049
3 长光卫星技术有限公司吉林省卫星遥感应用技术重点实验室, 吉林 长春 130102
摘要
提出了一种基于集成卷积神经网络(CNN)的遥感影像场景分类算法。通过构建反向传播网络实现了场景图像的复杂度度量;根据图像的复杂度级别,选择CNN对图像进行分类,完成了遥感影像的场景分类。使用所提出的算法对NWPU-RESISC45公开数据集进行了实验验证,取得了89.33%(第一类实验)和92.53%(第二类实验)的分类准确率,平均运行时间为0.41 s。相比于精调训练的VGG-16模型,所提算法的分类准确率分别提升了2.19%和2.17%,预测速率提升了33%,证明了其有效性和实用性。
Abstract
A scene classification algorithm of remote sensing images based on the integrated convolutional neural network (CNN) is proposed. A back-propagation network is constructed to measure the complexity of scene images. The classification of these images is conducted with the CNN based on the complexity level of each image, thus, the scene classification of remoting sensing images is achieved. With the proposed algorithm, the experimental verification of the open data of NWPU-RESISC45 is conducted and the classification accuracy of 89.33% for Type I test and that of 92.53% for Type II are obtained, respectively. The average running time is 0.41 s. Compared with the VGG-16 model for fine tuning and training, the classification accuracy by the proposed algorithm is increased by 2.19% and 2.17%, respectively. Simultaneously, the prediction rate is increased by 33%. Thus, the efficiency and practicality of this proposed algorithm are confirmed.

张晓男, 钟兴, 朱瑞飞, 高放, 张作省, 鲍松泽, 李竺强. 基于集成卷积神经网络的遥感影像场景分类[J]. 光学学报, 2018, 38(11): 1128001. Xiaonan Zhang, Xing Zhong, Ruifei Zhu, Fang Gao, Zuoxing Zhang, Songze Bao, Zhuqiang Li. Scene Classification of Remote Sensing Images Based on Integrated Convolutional Neural Networks[J]. Acta Optica Sinica, 2018, 38(11): 1128001.

本文已被 10 篇论文引用
被引统计数据来源于中国光学期刊网
引用该论文: TXT   |   EndNote

相关论文

加载中...

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!