圈养高山麝种群年龄组的粪便光谱判别

提出了一种利用粪便可见-近红外反射光谱进行圈养高山麝种群年龄组分析的新方法.以FieldSpec~((R))3地物光谱仪采集了145份高山麝粪便(成体麝粪样45份,亚成体和幼体各50份)的光谱数据,将其随机分成训练集(100份)和检验集(45份).光谱经S.Golay平滑和一阶导数处理后以主成分分析法(PCA)降维.以前6个主成分(含原始光谱95.00%的特征信息)作为新变量,利用训练集样本,分别以Fisher线性判别、Bayes逐步判别以及BP-神经网络(BP-ANN)3种方法建立高山麝种群年龄组的分析模型.对检验集45个未知样的预测表明,BP-ANN模型判别的准确率最高,为84.44%.3种方法所建的模型对幼麝粪样判别的准确率最高,可达93.33%.分析发现亚成体粪样具有过渡性质,但幼麝粪样与成体粪样易于判别.结果表明,利用粪便的可见-近红外反射光谱进行高山麝年龄组的快速、非接触性判别是可行的,且PCA 结合BP-ANN判别是一种优选方法.

Differentiation of population age groups of captive alpine musk deer (Moschus chrysogaster) based on the visible-near infrared reflectance spectra of their feces

A new method was developed to analyze the population age groups of the captive Alpine musk deer (Moschus chrysogaster) based on the visible-near infrared reflectance spectra of their feces. A total of 145 feces samples including 50, 50 and 45 from fawns, subadults and adults, respectively, were collected at Xinglongshan Musk Deer Farm, Gansu Province. The spectral scanning on the samples was carried out in a darkroom (18-22℃, 22%-25% humidity and a special halogen lamp as the sole light source) with an ASD FieldSpec~((R))3 spectrometer. The spectral data were assigned randomly to two sets for calibration (100 data) and validation (45 data) each. The data were pretreated by the methods of S. Golay smoothing and the first derivative. The pretreated spectra were analyzed by principal component analysis (PCA), and the resultant top 6 principal components, which accounted for 95.00% of the variation among original spectral data, were used as the inputs for modeling. Individual calibration models were developed for differentiation of the population age groups of Alpine musk deer by Fisher linear discriminant (FLD), Bayes stepwise discriminant (BSD) or Back-propagation artificial neural network (BP-ANN). The predicted outcomes from the validation models of the 45 sample spectral data indicated that the accuracy rates of prediction were 84.44%, 82.22% and 80.00% for BP-ANN, BSD and FLD, respectively. The highest differentiation accuracy of 93.33% was achieved in the fawns among the three age groups. The differences in the fecal spectra among the three age groups revealed the component diversity of the feces, which corresponded with the ontogenetic development of the digestive system of Alpine musk deer. The fecal samples of subadults were of transitional attributes, but those of fawns or adults were readily differentiated by the models. The results suggested that visible-near infrared reflectance spectra of feces was a rapid and non-invasive method for distinguishing the population age groups of Alpine musk deer, and the PCA combined with BP-ANN was a preferred method.