基于多参考相关系数法和BP-ANN建立紫石英的近红外光谱定性模型

目的:基于多参考相关系数法和误差反向传递人工神经网络(BP-ANN)建立矿物药紫石英的近红外光谱定性模型,用于紫石英的生品、煅制品、醋煅品和伪品的快速鉴别。方法:采集紫石英、紫石英煅制品、紫石英醋煅品及紫石英伪品这4类不同紫石英样品的近红外光谱,对光谱进行二阶导数和9点平滑预处理,计算多项相关系数。在Matlab 2014a软件中将多项相关系数作为输入数据,以BP-ANN建立4类不同紫石英样品的快速鉴别模型。结果:建立了紫石英近红外光谱BP-ANN鉴别模型,模型验证结果显示,15批验证样品中14批样品预测结果正确,仅1批样品预测有误,准确率达93.33%。结论:建立的紫石英近红外光谱BP-ANN鉴别模型能通过一次性整合的运算区分紫石英生品、煅制品、醋煅品及其伪品,鉴别结果准确可靠。此外,模型在近红外光谱相关系数法基础上,以多个参考光谱为对照计算所得的多组相关系数作为网络特征输入数据,实现了光谱数据的压缩。

Establishment of Near Infrared Spectral Qualitative Model of Fluoritum Based on Multi-reference Correlation Coefficient Method and BP-ANN

Objective: To establish a near infrared spectral qualitative model of Fluoritum for rapidly identifying raw,calcined,vinegar calcined products and its counterfeits based on multi-reference correlation coefficient method(MRCC) and back-propagation artificial neural networks(BP-ANN). Method: The near infrared spectra of 4 kinds of Fluoritum samples were collected,including raw products,calcined products,vinegar calcined products and counterfeit products.The second derivative and 9 points smoothing were used to calculate multiple correlation coefficients.Multiple correlation coefficients were used as input data in Matlab 2014a software,and a rapid identification model for 4 different types of Fluoritum samples was established by BP-ANN. Result: BP-ANN identification model of near infrared spectroscopy of Fluoritum was established.The prediction results of 14 batches of samples were correct,only 1 batch of samples was predicted error,the accuracy rate was 93.33%. Conclusion: The BP-ANN identification model of near infrared spectroscopy of Fluoritum can be used to distinguish among 4 different types of samples by a one-time integrated operation,and the result is accurate and reliable.In addition,based on the correlation coefficient method of near infrared spectroscopy,the multi-group correlation coefficient calculated by the multiple reference spectra is used as the input data of the network characteristics,the compression of spectral data is realized.