基于样本优化选取的光谱重建方法研究

Study on Spectral Reconstruction Method Based on Optimized Selected Samples

针对光谱反射率重建中已有样本选取方法的不足,提出了一种基于核模糊C聚类的样本优化选取方法。该方法综合考虑了光谱反射率空间的广泛性和色度空间的相似性,较大程度满足了光谱重建的精度。首先采用已有样本选取法在光谱反射率空间选取C个样本作为聚类初始点,再将原光谱转化到色度空间进行聚类,同时引入核函数将二维色度空间映射到三维特征空间,使得特征数据线性可分,从而达到更好的划分效果。实验结果表明,使用该方法选取训练样本进行光谱反射率重建能够进一步提高光谱重建精度,色度评价和光谱评价结果均好于已有方法。

Aiming at the defect of existing sample selection methods for spectral reflectance reconstruction, a new method based on the kernel fuzzy C-means clustering optimization sample selection is presented. This method considers the extensiveness of spectral reflectance space and the similarity of chroma space, meeting the accuracy of spectral reconstruction. Firstly, C samples are selected as the initial clustered core in the spectral reflectance space. And the original spectrais turned into （a＊, b＊） color space. At the same time, 2D chrominance space is mapped to the 3D feature space with kernel function, making characteristics linearly separable, to achieve a better result. The experiments show that this method selects training samples for spectral reflectance reconstruction can further improve the accuracy of spectral reconstruction, and it is better than the existing methods in the chroma and accuracy assessment effects.