用于中药质量整体辨识的小波基化学模式图像识别法

Chemical pattern image recognition method for identifying the quality of herbal medicines based on wavelet transform

提出一类基于小波变换的化学模式图像识别法 ,并应用于中药质量的直观表征与整体辨识 .根据受检样品的相关化学组分含量波动程度 ,通过不同尺度的离散小波变换或小波包变换来有效提取量测数据集中隐含的化学指纹特征 ,放大不同样品间化学模式差异 ;再采用二维灰度图对抽提出的化学特征信息进行图像表征 ,形成视觉可分辨的化学模式图像 ,从而实现中药质量的直观整体辨识 .将该方法用于不同等级、产地及采收年份的当归样品鉴定 ,结果表明该方法可有效地解决中药质量整体辨识难题 ,为化学模式识别方法研究开辟了新的途径 .

A chemical pattern image recognition method based on wavelet transform is proposed, and applied to represent and identify the quality of herbal medicines. Based on the discrepancy of chemical component content among analyzed samples, discrete wavelet or wavelet packets transform under different scales were selected to extract the fingerprint features and intensify the pattern difference. Then, the extracted feature information was represented with two dimensional grayscale images so that the quality of herbal medicine can be integrally identified with visual sense. This method was applied to classify DangGui samples from different quality grades, areas and years. The results showed that the different samples can be classified with variable resolution based on the chemical pattern image features. It was proved that the proposed method is a new effective tool for chemical pattern recognition.