摘要
研究基于珠海一号高光谱影像的冬小麦识别提取技术,提出基于多级融合的多时相高分辨率高光谱冬小麦提取方法。本文从珠海一号高光谱影像入手,利用高分辨率影像改善高光谱影像空间分辨率,通过主成分分析降维、多种特征提取技术,大幅减少计算量的同时提高分类精度,Kappa系数提升0.05。针对融合影像是否有效、高程特征如何正确使用、U型语义分割网络(U-Net)和深度卷积语义分割网络(DeepLab)如何选择等问题,文中以4个实验对比说明,验证了该方法可以有效改善分类结果。
A multi-temporal high-resolution hyperspectral winter wheat extraction method based on multi-level fusion was proposed by studying winter wheat identification and extraction technology based on Zhuhai-1 hyperspectral image.This paper starts with the hyperspectral image of Zhuhai No.1,and uses high-resolution images to improve the spatial resolution of hyperspectral images.Through Principal Components Analysis(PCA)dimensionality reduction and multiple feature extraction techniques,the calculation amount is greatly reduced while the classification accuracy is improved,and the kappa coefficient is increased by 0.0501.Aiming at the problems of whether the fusion image is effective,how to use the elevation features correctly,and how to select the U-Net and DeepLab semantic segmentation network,this paper compares four experiments results to verify that the method can effectively improve the classification results.
作者
郭欣怡
吕扬
王源
宣兆新
GUO Xinyi;LYU Yang;WANG Yuan;XUAN Zhaoxin(Beijing Institute of Surveying and Mapping,Beijing 100038,China;Beijing Key Laboratory of Urban Spatial Information Engineering,Beijing 100038,China)
出处
《北京测绘》
2023年第10期1391-1396,共6页
Beijing Surveying and Mapping
基金
北京市自然科学基金(8222011)。