光谱学与光谱分析, 2017, 37 (6): 1718, 网络出版: 2017-07-10   

基于近红外高光谱技术和特征波谱分析方法的竹类判别研究

Discriminant Analysis of Bamboo Leaf Types with NIR Coupled with Characteristic Wavelengths
作者单位
浙江大学生物系统工程与食品科学学院, 浙江 杭州 310058
摘要
竹叶含有丰富的功能性成分, 具有良好的抗氧化、 调节血脂、 抗癌、 保护心脑血管等功效, 在食品和药品等领域具有较高的应用价值, 但不同品种的竹叶其功能性成分差异较大。 传统对于竹类品种的鉴别主要是通过观察竹叶大小、 纹理、 竹枝分枝和竹竿高度等, 效率低且错误率较高, 因此, 快速准确的区分不同品种的竹叶, 是竹类资源开发和加工过程中的重要任务之一。 采用近红外高光谱(900~1 700 nm)技术对我国不同产地的12种竹叶进行鉴别分析。 用主成分分析(PCA)对竹叶进行聚类分析, 应用主成分因子中X-loading(XL)和random frog(RF)算法进行特征波段的提取, 分别得到6条(931, 945, 1 217, 1 318, 1 473和1 653 nm)和12条(1 052, 1 140, 1 163, 1 177, 1 180, 1 193, 1 230, 1 241, 1 477, 1 483, 1 629和1 649 nm)特征波段, 并基于全波段(238条波长)及采用以上算法所得的特征波段建立最小二乘-支持向量机(LS-SVM)判别分析模型, 其识别率分别为9917%(全波段), 9583%(XL算法), 9583%(RF算法)。 最后, 采用受试者工作特征曲线(receiver operating characteristic curve, ROC curve)对LS-SVM模型的判别效果进行验证, 结果表明, 曲线下面积(AUC)均在098以上, 说明近红外高光谱结合LS-SVM可以很好地实现竹类的鉴别分析, 这为竹叶的食用和药用价值的开发利用提供理论参考。
Abstract
Bamboo leaves are rich in many kinds of biological active components such as flavonoid, phenolic acid and polysaccharide, which have demonstrated good effects of anti-oxidant, blood lipid regulation, anti-cancer, cardiovascular and cerebrovascular protection etc. However, the content of active constituents exhibits great differences among different bamboo species. The traditional identification of bamboo species is mainly based on the observation of bamboo leaf size, bamboo texture and branching height etc. which has the disadvantages of low efficiency and high error rate. Therefore, to distinguish the varieties of bamboo with a rapid and accurate method is an important task in the development and utilization of bamboo leaves. A near-infrared (900~1 700 nm) hyperspectral technique was used to identify 12 bamboo species from different regions of China. Principal component analysis (PCA) was applied to make the cluster analysis. PCA X-loading (XL) and Random Frog (RF) algorithm was chosen to extract spectral feature and 6 characteristic wavelengths (931, 945, 1 217, 1 318, 1 473 and 1 653 nm) and 12 characteristic wavelengths (1 052, 1 140, 1 163, 1 177, 1 180, 1 193, 1 230, 1 241, 1 477, 1 483, 1 629 and 1 649 nm) were selected respectively. Then, the spectra based on the selected wavelengths were set as the input values of Least Squares-Support Vector Machine (LS-SVM) model to perform the discriminant analysis. At last, the properties of the three LS-SVM models were evaluated with Receiver Operating Characteristic curve (ROC curve). Results showed that (1) in the three LS-SVM models, the recognition rate of full band, XL algorithm and RF algorithm were 9917%, 9583% and 9583% respectively, (2) the area under the curve (AUC) in ROC curve were all reach over 098. In conclusion, the bamboo leaves from different regions could be identified by near-infrared hyperspectral technique combined with chemometrics methods, which provided a theoretical foundation for efficient utilization of bamboo leaves.

楚秉泉, 赵艳茹, 何勇. 基于近红外高光谱技术和特征波谱分析方法的竹类判别研究[J]. 光谱学与光谱分析, 2017, 37(6): 1718. CHU Bing-quan, ZHAO Yan-ru, HE Yong. Discriminant Analysis of Bamboo Leaf Types with NIR Coupled with Characteristic Wavelengths[J]. Spectroscopy and Spectral Analysis, 2017, 37(6): 1718.

本文已被 1 篇论文引用
被引统计数据来源于中国光学期刊网
引用该论文: TXT   |   EndNote

相关论文

加载中...

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!