摘要
针对高分二号遥感影像分类精度低的问题,该文提出了一种基于加权空-谱局部保持投影的高分二号遥感影像分类方法。该方法首先采用加权空-谱算法对高分二号遥感影像进行几何空间重构处理,获取新的影像数据;然后利用融合LPNPE方法与LPP方法构建新的投影矩阵,充分利用了样本数据光谱类别信息又兼顾局部相邻空间的几何相互关系;最后,将投影后的数据作为SVM分类器的输入,对高分二号遥感影像进行分类。实验结果表明,该方法分类精度为97.47%,Kappa系数为0.9753,能有效识别地物边界。
Aiming at the problem of low classification accuracy of Gaogao-2 remote sensing image,a classification method of Gaogao-2 remote sensing image based on weighted space-spectrum local preserving projection is proposed in this paper.Firstly,the weighted space-spectrum algorithm is used to reconstruct the geometric space of Gaogao-2 remote sensing image to obtain new image data;Then,the fusion LPNPE method and LPP method are used to construct a new projection matrix,which make full use of the spectral category information of the sample data and take into account the geometric relationship of local adjacent spaces;Finally,the projected data is used as the input of SVM classifier for high score No.2 remote sensing image classification.The experimental results show that the classification accuracy of this method is 97.47%and the Kappa coefficient is 0.9753,which can effectively distinguish the boundary of ground objects.
作者
何晓琳
He Xiaolin(Qinghai Natural Resources Comprehensive Investigation and Monitoring Institute)
出处
《勘察科学技术》
2022年第1期19-22,共4页
Site Investigation Science and Technology
关键词
空间几何特征
光谱特征
加权空-谱局部保持投影
高分二号影像分类
spatial geometric feature
spectral feature
weighted space-spectral local preserving projection
image classification of Gaofen-2
