基于数学形态学的高光谱图像降维方法

Dimensionality reduction method of hyperspectral image based on mathematical morphology

高光谱图像的Kappa系数较低,降维存在困难,设计一种基于数学形态学的高光谱图像降维方法。通过色彩空间变换实施数据的阴影检测,完成转换色彩空间与提取暗区,再根据数学形态学去除噪声。处理高光谱图像的色阶。完成色阶处理后需要实施阴影恢复操作。设计局部保留投影算法完成近邻关系图的构造,确定权重矩阵,分解特征值,最后通过低维描述计算实现高光谱图像降维处理。在实验中,使用了2个高光谱图像数据集,验证设计方法的性能。实验结果表明,设计方法的整体分类精度高于现有降维方法,同时Kappa系数也高于现有降维方法,实现了性能的大幅提升。

The kappa coefficient of hyperspectral image is low and dimensionality reduction is difficult.A dimensionality reduction method of hyperspectral image based on mathematical morphology is designed.The shadow detection of data is implemented through color space transformation,the color space is transformed and the dark area is extracted,and then the noise is removed according to mathematical morphology.Process the color scale of hyperspectral image.Shadow recovery is required after color scale processing.A local preserving projection algorithm is designed to construct the nearest neighbor graph,determine the weight matrix,decompose the eigenvalues,and finally reduce the dimension of hyperspectral image through low dimensional description calculation.In the experiment,two hyperspectral image data sets are used to verify the performance of the design method.The experimental results show that the overall classification accuracy of the design method is higher than the existing dimensionality reduction methods,and the kappa coefficient is also higher than the existing dimensionality reduction methods,which greatly improves the performance.