采用训练样本排序框定初始胞元、差异优先搜索最优胞元和初始胞元再扩大方法,建立了CYNSN 模型的胞元搜索算法。实验结果表明,CYNSN 模型胞元搜索算法能使CYNSN 反向模型在色度和光谱反射率上实现高精度复制,和已有的CYNSN 模型相比,胞元搜索算法不仅保持了CYNSN 模型的精度,而且效率提高了10倍。

A high-efficient cell-search method for CYNSN model was proposed, which included framing initial cell by ordering the training sample set, searching optimal cell using the priority in difference method step by step, and expanding initial cell. Experimental result showed that the proposed cell-search method can make the backward CYNSN model copy accurately in spectral and colorimetric reflectance; compared with existing CYNSN model, the CYNSN model using the proposed cell-search method improves efficiency more than 10 times with the same precision.