| 不同时段刺五加叶片黄酮含量的高光谱无损估测 |
| |
| 引用本文： | 王树文,修成,董元,姚煜,李晓峰,李雷,刘珺.不同时段刺五加叶片黄酮含量的高光谱无损估测[J].农机化研究,2022,44(4):119-126,268. |
| |
| 作者姓名： | 王树文 修成 董元 姚煜 李晓峰 李雷 刘珺 |
| |
| 作者单位： | 岭南师范学院, 广东 湛江 524048;东北农业大学 电气与信息学院, 哈尔滨 150030;东北农业大学 电气与信息学院, 哈尔滨 150030;中国科学院 东北地理与农业生态研究所, 长春 130102;岭南师范学院, 广东 湛江 524048;中国科学院 东北地理与农业生态研究所, 长春 130102;中国科学院 东北地理与农业生态研究所, 长春 130102;中国科学院大学, 北京 100049;国网哈尔滨供电公司, 哈尔滨 150010 |
| |
| 基金项目： | 广东省教育厅特色创新项目(2018KTSCX129);岭南师范学院人才专项(ZL2007)。 |
| |
| 摘 要： | 针对野生刺五加叶片中黄酮含量的测量方法繁琐、时间较长及需破坏叶片等问题,提出了一种基于高光谱技术对不同时段的刺五加叶片中黄酮含量的估算模型。首先,分析提取地域、年龄、长势相近的20株刺五加叶片光谱特征,通过对叶片进行烘干、磨粉及利用紫外分光光度计等化学方法测得叶片中黄酮的真实含量,并选择4种预处理互相结合、比较的方式,判断出最优预处理模型;通过SPA与PCA算法的结合,选择出较明显的特征波段,通过MatLab2018a将特征波段的反射率分别与40组预测集验证相关性后,再分别选取预测值和20组实测值与BP神经网络、支持向量机进行模型建立。实验结果表明:利用BP神经网络建立的模型的校正集决定系数Rc2分别为0.8649、0.7976、0.8485,支持向量机建立的模型的校正集决定系数Rc2分别为0.7526、0.7742、0.7243,证明SNV和1 Der结合的预处理方式与BP神经网络所构建的模型效果最好。研究为高光谱技术对刺五加叶片中黄酮的反演提供了有力的支持,也会提高工业和药用采摘的效率及刺五加的利用价值。
|
| 关 键 词： | 刺五加 黄酮 光谱预测 主成分分析 BP神经网络 |
Non-destructive Estimation of Flavonoids in Leaves of Acanthopanax Senticosus in Different Periods |
| |
| Wang Shuwen,Xiu Cheng,Dong Yuan,Yao Yu,Li Xiaofeng,Li Lei,Liu Jun.Non-destructive Estimation of Flavonoids in Leaves of Acanthopanax Senticosus in Different Periods[J].Journal of Agricultural Mechanization Research,2022,44(4):119-126,268. |
| |
| Authors: | Wang Shuwen Xiu Cheng Dong Yuan Yao Yu Li Xiaofeng Li Lei Liu Jun |
| |
| Institution: | (Lingnan Normal College, Zhanjiang 524048,China;College of Electrical and Information,Northeast Agricultural University, Harbin 150030,China;Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences, Changchun 130102,China;Chinese Academy of Sciences, Beijing 100049,China;Harbin Power Supply Company,Harbin 150010,China) |
| |
| Abstract: | Aiming at the problem that the method of measuring flavonoids in wild Acanthopanax senticosus leaves is cumbersome,the time is longer and the leaf needs to be destroyed.In this paper,a hyperspectral model for estimating the content of flavonoids in Acanthopanax senticosus leaves was proposed.The spectral characteristics of the leaves were analyzed.The true content of flavonoids in the leaves was measured by drying,grinding and ultraviolet spectrophotometer.The results showed that the true content of flavonoids in the leaves was mixed by smoothing(SG),multiple scattering correction(MSC/EMSC),standard normal transformation(SNV)and first derivative(1Der)the original spectrum was preprocessed in a combined manner.after comparison,snv and 1der were selected to combine each other as the final preprocessing method.through the combination of spa and pca algorithm,the characteristic band was selected and the principal component of pc1 and pc2 total sum of 97%was selected.after verifying the correlation of the reflectivity of the characteristic band with the 40 groups of prediction sets by matlab2018a,then select the predicted values and 20 groups of measured values to model with bp neural network and support vector machine,respectively.The experimental results show that the correction set coefficient R of the model based on BP neural network R 2 c at 0.8649,0.7976,0.8485,respectively.the model established by support vector machine has a correction set determination coefficient of R 2 c of 0.7526,0.7742,0.7243,respectively.the modeling results show that the model constructed by combining snv and 1der with bp neural network is the best.At the same time,it also provides a more powerful support for the inversion of flavonoids in the leaves of Acanthopanax senticosus,which will also improve the efficiency of industrial and medicinal picking,and further improve the utilization value of Acanthopanax senticosus. |
| |
| Keywords: | Acanthopanax flavonoids spectral prediction principal component analysis BP neural network |
| 本文献已被 维普 万方数据 等数据库收录！ |
|
