基于模糊聚类分析和曲线拟合的FTIR不同产地鸡骨草、毛鸡骨草鉴别分析

利用傅里叶变换红外光谱技术结合模糊聚类分析法和曲线拟合对鸡骨草、毛鸡骨草进行产地鉴别。选用曼哈顿距离单位计算的相异度聚类分析结果最优,5个产地鸡骨草都可完全区分开来,毛鸡骨草则只能鉴别出3个产地,南宁和钦州2个产地发生重叠;为了进一步鉴别不同产地鸡骨草和毛鸡骨草,对鸡骨草和毛鸡骨草平均红外谱图中1 034 cm-1吸收峰进行曲线拟合,野生鸡骨草拟合出11个子峰,其他各产地鸡骨草均只能拟合出9个子峰,上林产毛鸡骨草拟合出9个子峰,其他各产地毛鸡骨草均只能拟合出8个子峰,而且不同产地鸡骨草和毛鸡骨草拟合出的子峰位置和归一化强度都不完全相同;模糊聚类分析法和曲线拟合法有机结合能够使产地鉴别结果更加准确。

Origins Determination of Herba Abri cantoniensis and Herba Abri mollis with FTIR Combined with Fuzzy Cluster and Curve-Fitting

The methods of fuzzy cluster and curve-fitting combined with FTIR were used to determine the origins of Herba Abri cantoniensis and Herba Abri mollis. The spectra of Herba Abri cantoniensis and Herba Abri mollis are similar, both with typi-cal spectral shapes. The two spectra can be divided into 3 parts: the 1st is 3 500-2 800 cm~(-1) , containing stretching hands of -OH, N-H, and CH_2; the 2nd is 1 800-800 cm~(-1), containing stretching hands of ester carbonyl group and indican C-O(H), vibrational bands of C-C and benzene ring; The 3rd is 800-400 cm~(-1), containing skeletal vibration and scissoring vibration of molecular. The recorded FTIR spectral data were processed by 9-point-smoothing, 1st derivative, SNV and fuzzy cluster analy-sis sequentially. The fuzzy cluster analysis was carried out by similarity or dissimilarity matrix, and two matrices are computed with Manhattan and Euclidean distance. The results indicated that the optimization used Manhattan and dissimilarity matrix, and 5 origins of Herba Abri cantoniensis were perfectly discriminated, but 2 origins of Herba Abri mollis were mixed and identified from the other 3 origins. So the characterized bands at 1 034 cm~(-1) of the average 1-D spectra of Herba Abri cantoniensis and Herba Abri mollis were fitted combining 2nd derivative for further distinguishing their spectral characteristic. The results of curve-fitting showed that the bands of wild HerbaAbri cantoniensis and the other origin ones were decomposed to 11 and 9 com-ponent bands respectively, but the bands of Shanglin and the other origins Herba Abri mollis were decomposed to 9 and 8 com-ponent bands dissimilarly, and the locations and normalized densities of these component hands were different. From this, to-gether with the results of fuzzy cluster analysis, it is concluded that the combination of two methods may identify the origins of Herba Abri cantoniensis and Herba Abri mollis availably.