摘要
提出一种基于堆叠降噪自编码器的影像特征融合变化检测方法。首先在影像分割对象的像素级特征和对象级特征分析、设计与提取基础上,通过像素级特征主成分分析与特征优选处理算法,实现遥感影像多类型特征融合并构建分割对象的多维特征向量;其次利用堆叠降噪自编码器的高维复杂数据处理能力,实现遥感影像多维融合特征的变化检测。实验结果表明,该方法与传统的像素级或对象级特征变化检测方法相比,具有更高的变化检测精度。
An image feature fusion change detection method based on stacked denoising auto-encoder is proposed in the paper. Firstly, pixel level features of segmented images are optimized and selected by principal component analysis, and they are combined with object level features to construct multi-dimensional fusion features. Then, the stacked denoising auto-encoder is used to realize the change detection for its good performance in high-dimensional complex data processing. The experimental results show that the proposed method has higher change detection accuracy than the traditional pixel level or object level feature change detection method.
作者
史文洁
戴晨光
赵莹
王岩
王志坚
SHI Wenjie;DAI Chenguang;ZHAO Ying;WANG Yan;WANG Zhijian(Information Engineering University,Zhengzhou 450001,China)
出处
《测绘科学技术学报》
CSCD
北大核心
2021年第4期391-397,共7页
Journal of Geomatics Science and Technology
基金
国防科工局高分专项(42-Y30B04-9001-19121)。
关键词
堆叠降噪自编码器
特征融合
变化检测
遥感影像
深度学习
stacked denoising auto-encoder
feature fusion
change detection
remote sensing image
deep learning
