摘要
遥感影像分类是遥感技术应用的一个重要环节;机器学习和深度学习能够实现精确、自动化、迅速、可定义和规模化的遥感影像分类。本文选取机器学习算法支持向量机和深度学习算法卷积神经网络、深度置信网络、栈式自编码网络共计4种分类算法进行对比研究,并对支持向量机核函数的参数以及深度学习算法的神经元数量开展寻优以到达最高分类精度。实验结果表明,深度学习算法栈式自编码网络的总体分类精度最高,分类效果最好,在地物复杂多样地区开展遥感地物分类时具有较好的适用性和推广价值。
Remote sensing image classification is an important part of the application of remote sensing technology.Machine learning and deep learning can achieve accurate,automated,rapid,definable,and scalable remote sensing image classification.This article compares four classification algorithms,namely machine learning algorithm support vector machine,deep learning algorithm convolu-tional neural network,deep confidence network,and stack based self coding network,and optimizes the parameters of support vector machine kernel function and the number of neurons in deep learning algorithm to achieve the highest classification accuracy.The ex-perimental results show that the overall classification accuracy and performance of the deep learning algorithm stack based self coding network are the highest,and it has good applicability and promotion value in remote sensing land classification in complex and diverse areas.
作者
陈香
CHEN Xiang(Tianjin Surveying and Mapping Institute Co.,Ltd.,Tianjin 300381,China)
出处
《测绘与空间地理信息》
2024年第7期72-75,共4页
Geomatics & Spatial Information Technology
关键词
遥感影像分类
支持向量机
卷积神经网络
深度置信网络
栈式自编码网络
remote sensing image classification
support vector machine
convolutional neural network
deep confidence network
stacked self coding network
