中文摘要：

本研究用变形雅可比（p=4，q=3）-傅立叶矩（PJFM’s），对草红花粉末涂片的花粉粒、分泌管碎片、花瓣细胞、花药中部细胞、花瓣顶端细胞、花柱头细胞等六个显微结构原始照片进行特征提取，发现不同特征点的不变矩值有着明显的区别，相同显微特征点不同方位、不同灰度、不同大小的变形图像的不变矩值几乎相同。并且通过重建实验，证明使用很少的PJFM’s就能很好地表达原始图像，且PJFM’s具有较强的自动抗噪声能力。PJFM’s可作为一种高度浓缩的图像特征，用于中药粉末显微特征快速自动识别系统的特征提取量。该研究结果为中药材显微识别和鉴定的数字化提供重要的技术基础。

英文摘要：

Pseudo-Jacobi（ p = 4, q = 3）-Fourier Moments （ PJFM＇ s） were applied to extract image featurs from six microscopic characteristic original photos of flos carthami, which included farina, fragment of secretory pipe, top cell of petal, petal cell, stigma cell, middle cell of anther in this paper. The experimental result indicated that the values of invariant moment were very different for different microscopic characteristic images and were nearly same for different versions of a microscopic characteristic image. The quality of reconstructed images also proved that PJFM＇ s were the highly condensed feature descriptor for image digitalization, and PJFM＇ s were also automatically robust to noise. This study may provide some technical basis for digitalization and visualization of Chinese herbal identification.