基于贝叶斯优化卷积神经网络的金线莲伪品鉴别

doi:10.3969/j.issn.1004-1524.2022.02.21

浙江农业学报 ›› 2022, Vol. 34 ›› Issue (2): 391-396.DOI: 10.3969/j.issn.1004-1524.2022.02.21

基于贝叶斯优化卷积神经网络的金线莲伪品鉴别

柴琴琴¹^,²^,³(), 曾建¹^,²^,³, 张勋⁴

1.福州大学电气工程与自动化学院,福建福州 350108
2.福州大学教育部医疗器械与制药技术重点实验室,福建福州 350108
3.福州大学晋江科教园,福建晋江 362251
4.福建中医药大学药学院,福建福州 350122

收稿日期:2020-09-18 出版日期:2022-02-25 发布日期:2022-03-02
作者简介:柴琴琴(1985—),湖北荆门人,女,博士,副教授,研究方向为化工建模优化控制、模式识别算法研究。E-mail: qq.chai@fzu.edu.cn
基金资助:
国家自然科学基金(61773124);福州大学晋江科教园科研项目(2019-JJFDKY-48)

Identification of fake Anoectochilus roxburghii based on Bayesian optimized convolutional neural network

CHAI Qinqin¹^,²^,³(), ZENG Jian¹^,²^,³, ZHANG Xun⁴

1. College of Electrical Engineering and Automation, Fuzhou University, Fuzhou 350108, China
2. Ministry of Education Key Laboratory of Medical Instrument and Pharmaceutical Technology, Fuzhou University, Fuzhou 350108, China
3. Jinjiang Science and Education Park of Fuzhou University, Jinjiang 362251, Fujian, China
4. School of Pharmacy, Fujian University of Traditional Chinese Medicine, Fuzhou 350122, China

Received:2020-09-18 Online:2022-02-25 Published:2022-03-02

摘要/Abstract

摘要：

在金线莲粉末中掺入同科台湾银线莲或斑叶兰、血叶兰粉末的现象严重影响了金线莲药材的药效和市场秩序,寻找一种快速有效的方法来鉴别掺假金线莲是亟待解决的问题。针对传统鉴别方法特征提取自适应性的不足以及卷积神经网模型结构复杂、超参数难以调节的难点,本文提出基于一维卷积神经网络的掺假金线莲鉴别模型,并利用贝叶斯优化算法优化卷积神经网络超参数,实现了超参数自动优化调节。实验结果表明,经过超参数寻优后的卷积神经网络相比传统机器学习模型更有竞争力,所提出的基于贝叶斯优化的一维卷积神经网络模型可以快速有效地鉴别金线莲及其伪品。

关键词: 卷积神经网络, 金线莲, 定性分析, 化学计量学方法, 贝叶斯优化

Abstract:

The phenomenon of mixing the same family of Anoectochilus roxburghii(A. roxburghii ) powder with the genus Anoectochilus formosanus or Goodyera schlechtendaliana, Ludisia discolor has seriously affected the drug efficacy and market order of A. roxburghii. Therefore, finding a fast and effective method to identify adulterated A. roxburghii is an urgent problem to be solved. In view of the shortcomings of the adaptive feature extraction of traditional identification methods and the difficulties of complex structure and difficult adjustment of super parameters of convolution neural network model, an adulterated A. roxburghii identification model based on 1D convolutional neural network (1D-CNN) was proposed in this paper, in which the Bayesian optimization algorithm was proposed to optimize its hyperparameters to realize automatic optimization and adjustment of the hyperparameters. Experimental results showed that the 1D-CNN model with hyperparameter optimization was more competitive than other traditional machine learning models. The proposed 1D-CNN model based on Bayesian optimization can quickly and effectively identify A. roxburghii and its counterfeits.

Key words: convolutional neural network, Anoectochilus roxburghii, qualitative analysis, chemometrics, Bayesian optimization

中图分类号:

TP183

柴琴琴, 曾建, 张勋. 基于贝叶斯优化卷积神经网络的金线莲伪品鉴别[J]. 浙江农业学报, 2022, 34(2): 391-396.

CHAI Qinqin, ZENG Jian, ZHANG Xun. Identification of fake Anoectochilus roxburghii based on Bayesian optimized convolutional neural network[J]. Acta Agriculturae Zhejiangensis, 2022, 34(2): 391-396.

图/表 6

图1 金线莲及其伪品平均近红外光谱图 A1,金线莲;A2,金线莲掺杂斑叶兰;A3,金线莲掺杂血叶兰;A4,台湾银线莲。下同。

Fig.1 Average near infrared spectra of A. roxburghii and its counterfeits A1,Anoectochilus roxburghii; A2, Anoectochilus roxburghii combined with Goodyera schlechtendaliana; A3, Anoectochilus roxburghii combined with Ludisia discolor; A4, Anoectochilus formosanus. The same as below.

表1 金线莲及其伪品样本信息

Table 1 Sample details of A. roxburghii and its counterfeits

样本类别 Category	样本数量 Number of samples	训练集 Training set	验证集 Validation set	预测集 Prediction set
A1	80	45	15	20
A2	84	47	16	21
A3	84	47	16	21
A4	80	45	15	20
总计 Total	328	184	62	82

图2 1D-CNN结构图

Fig.2 1D-CNN model structure

表2 1D-CNN、SVM和KNN超参数寻优设置

Table 2 1D-CNN, SVM and KNN hyperparameter optimization setting

模型 Model	超参数 Hyperparameter	搜索空间 Search space
1D-CNN	Filters	{12,16,32,64}
	Kernel-size	{3,5,7,15}
	Max pooling-size	{2,3,5}
	Units	{100,200,500}
	Batch-size	{12,16,32,64}
	Epoch	{100,150,200}
SVM	C	[0.001,1 000]
	Kernel	{rbf, linear, sigmoid, poly}
KNN	Neighbors	[1,50]
	Metric	{minkowski, euclidean,Chebyshev, manhattan}

图3 1D-CNN模型超参数寻优结果

Fig.3 1D-CNN model hyperparameter optimization results

表3 KNN、SVM和1D-CNN鉴别性能对比

Table 3 Performance comparisons of KNN, SVM and 1D-CNN %

模型Model	CIRT	CIRV	CIRP
KNN-B	95.65 d	95.16 c	90.24 b
SVM-B	91.30 c	93.54 d	84.15 d
1D-CNN	96.74 b	96.77 b	90.24 b
1D-CNN-B	98.37 a	100 a	98.78 a

参考文献 17

[1]	ZENG B Y, SU M H, CHEN Q X, et al. Antioxidant and hepatoprotective activities of polysaccharides from Anoectochilus roxburghii[J]. Carbohydrate Polymers, 2016, 153:391-398. DOI URL
[2]	YE S Y, SHAO Q S, ZHANG A L. Anoectochilus roxburghii: a review of its phytochemistry, pharmacology, and clinical applications[J]. Journal of Ethnopharmacology, 2017, 209:184-202. DOI URL
[3]	曹奕鸯, 李永清, 江金兰, 等. 不同种源金线兰及近缘种多糖和总黄酮含量的研究[J]. 福建农业学报, 2016, 31(6):604-610.
	CAO Y Y, LI Y Q, JIANG J L, et al. Polysaccharide and flavonoid contents in Anoectochilus roxburghii and related species[J]. Fujian Journal of Agricultural Sciences, 2016, 31(6):604-610.(in Chinese with English abstract)
[4]	TANG T T, DUAN X Y, KE Y, et al. Antidiabetic activities of polysaccharides from Anoectochilus roxburghii and Anoectochilus formosanus in STZ-induced diabetic mice[J]. International Journal of Biological Macromolecules, 2018, 112:882-888. DOI URL
[5]	王海阁, 许文, 张勋, 等. 林下栽培金线莲的生药鉴别[J]. 中药材, 2020, 43(2):303-308.
	WANG H G, XU W, ZHANG X, et al. Pharmacognostical identification of Anoectochilus roxburghii cultivated under forests[J]. Journal of Chinese Medicinal Materials, 2020, 43(2):303-308.(in Chinese)
[6]	陈莹, 任丽, 严桂杰, 等. 不同来源金线莲的HPLC指纹图谱[J]. 沈阳药科大学学报, 2019, 36(9):794-804.
	CHEN Y, REN L, YAN G J, et al. Fingerprint analysis of Anoectochilus roxburghii from different sources by HPLC[J]. Journal of Shenyang Pharmaceutical University, 2019, 36(9):794-804.(in Chinese with English abstract)
[7]	黄颖桢, 陈菁瑛, 赵云青, 等. 金线莲遗传多样性的ISSR分析[J]. 中草药, 2014, 45(15):2230-2234.
	HUANG Y Z, CHEN J Y, ZHAO Y Q, et al. ISSR analysis on genetic diversity of Anoectochilus roxburghii[J]. Chinese Traditional and Herbal Drugs, 2014, 45(15):2230-2234.(in Chinese with English abstract)
[8]	LI S L, WANG Z A, SHAO Q S, et al. Rapid detection of adulteration in Anoectochilus roxburghii by near-infrared spectroscopy coupled with chemometric methods[J]. Journal of Food Science and Technology, 2018, 55(9):3518-3525. DOI URL
[9]	郭云香, 陈龙, 李晓瑾, 等. 基于NIRS技术和PCA-SVM算法快速鉴别国产和进口啤酒花[J]. 药学实践杂志, 2019, 37(4):322-331.
	GUO Y X, CHEN L, LI X J, et al. Rapid identification of domestic and imported hops based on NIRS technology and PCA-SVM algorithm[J]. Journal of Pharmaceutical Practice, 2019, 37(4):322-331.(in Chinese with English abstract)
[10]	张丙芳, 王玉林, 刘成海, 等. 压榨及浸出工艺对大豆油稳定性影响及IPLS-SPA特征波段选择方法应用[J]. 食品科学, 2020, 41(10):8-13.
	ZHANG B F, WANG Y L, LIU C H, et al. Effects of different processing technologies on the stability of soybean oil and application of characteristic band selection by interval partial least squares and successive projections algorithm for near-infrared spectroscopic prediction of acid and carbonyl values[J]. Food Science, 2020, 41(10):8-13.(in Chinese with English abstract)
[11]	KRIZHEVSKY A, SUTSKEVER I, HINTON G E. ImageNet classification with deep convolutional neural networks[J]. Communications of the ACM, 2017, 60(6):84-90. DOI URL
[12]	REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6):1137-1149. DOI URL
[13]	CHEN X Y, CHAI Q Q, LIN N, et al. 1D convolutional neural network for the discrimination of aristolochic acids and their analogues based on near-infrared spectroscopy[J]. Analytical Methods, 2019, 11(40):5118-5125. DOI URL
[14]	ROCCO I, ARANDJELOVIC R, SIVIC J. Convolutional neural network architecture for geometric matching[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 41(11):2553-2567. DOI URL
[15]	MISHKIN D, SERGIEVSKIY N, MATAS J. Systematic evaluation of convolution neural network advances on the Imagenet[J]. Computer Vision and Image Understanding, 2017, 161:11-19. DOI URL
[16]	崔佳旭, 杨博. 贝叶斯优化方法和应用综述[J]. 软件学报, 2018, 29(10):3068-3090.
	CUI J X, YANG B. Survey on Bayesian optimization methodology and applications[J]. Journal of Software, 2018, 29(10):3068-3090.(in Chinese with English abstract)
[17]	SHAHRIARI B, SWERSKY K, WANG Z Y, et al. Taking the human out of the loop: a review of Bayesian optimization[J]. Proceedings of the IEEE, 2016, 104(1):148-175. DOI URL

基于贝叶斯优化卷积神经网络的金线莲伪品鉴别

Identification of fake Anoectochilus roxburghii based on Bayesian optimized convolutional neural network

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 6

参考文献 17

相关文章 8

编辑推荐

Metrics

本文评价

[1]	张晴晴, 刘连忠, 宁井铭, 吴国栋, 江朝晖, 李孟杰, 李栋梁. 基于YOLOV3优化模型的复杂场景下茶树嫩芽识别[J]. 浙江农业学报, 2021, 33(9): 1740-1747.
[2]	鲍烈, 王曼韬, 刘江川, 文波, 明月. 基于卷积神经网络的小麦产量预估方法[J]. 浙江农业学报, 2020, 32(12): 2244-2252.
[3]	蒋丰千, 李旸, 余大为, 孙敏, 张恩宝. 基于Caffe卷积神经网络的大豆病害检测系统[J]. 浙江农业学报, 2019, 31(7): 1177-1183.
[4]	金秀, 卢杰, 傅运之, 王帅, 许高健, 李绍稳. 基于深度卷积神经网络的小麦赤霉病高光谱病症点分类方法[J]. 浙江农业学报, 2019, 31(2): 315-325.
[5]	张善文, 黄文准, 尤著宏. 基于物联网和深度卷积神经网络的冬枣病害识别方法[J]. 浙江农业学报, 2017, 29(11): 1868-1874.
[6]	王丽芳;黄建;林霞*. 秋水仙素离体诱导金线莲多倍体的研究[J]. , 2010, 22(6): 760-763.
[7]	俞琦英;赵伟明. 油菜区域试验产量性状及其稳定性分析[J]. , 2010, 22(3): 0-340.
[8]	石建尧;王桂跃. 糯玉米区域试验非平衡数据产量性状及稳定性分析[J]. , 2008, 20(2): 0-134.