Research on Classification Method of Micro Character of Paris Polyphylla with Efficient Attention Module

doi:10.3778/j.issn.1002-8331.2011-0407

Abstract

Abstract: Paris polyphylla is an important component of many Chinese patent medicines. Different kinds of paris polyphylla are very similar in shape but different in quality, making it difficult to identify. In response to this problem, this paper uses a stereo microscope to collect two kinds of cross-sectional microscopic images of fresh rhizome of the paris polyphylla to identify them. Based on the ResNeXt101 model, it proposes the ES-Net module combining an efficient channel attention network and a spatial attention module. The output of each part of the ResNeXt module is further input into the ES-Net module. The Mixup method is used for data enhancement. The experimental results show that the highest accuracy of the classification of the two kinds of the paris polyphylla images is 94.95%, which is 2.07 percentage points higher than the original ResNeXt101 model. The results show that the ES-Net module can effectively enhance the characteristic information extraction ability of the original model ResNeXt101 on the cross-sectional microscopic image of fresh rhizome of the Paris polyphylla. Therefore, it improves the classification accuracy for the paris polyphylla. The deep learning method proposed has practical value for automatic identification of fresh rhizome of the paris polyphylla.

Key words: paris polyphylla, deep learning, attention module, ResNeXt-101 model, Mixup

摘要： 重楼是多种中成药的重要原料药材，重楼属植物不同品种药材的形状相似而品质不一，难以鉴别。针对该问题，通过体视显微镜采集滇重楼及毛重楼的新鲜根茎横切面显微图像进行鉴别。基于ResNeXt101模型，提出了结合高效通道注意力网络与空间注意力机制的ES-Net模块，将每部分ResNeXt模块的输出进一步输入到ES-Net模块中，并使用Mixup方法进行数据增强。实验结果显示改进的模型对两种重楼根茎横切面图像的分类精度最高，为94.95%，比原ResNeXt101模型提高了2.07个百分点。实验结果表明，提出的ES-Net模块能有效增强原模型ResNeXt101对重楼新鲜根茎横切面图像的特征提取能力，提高对其的分类精度，提出的深度学习方法对重楼新鲜根茎自动鉴别具有实用价值。

关键词: 重楼, 深度学习, 注意力机制, ResNeXt101模型, Mixup

LUO Xudong, LI Zhonggui, ZHANG Junhua, LI Xuefang, YU Wentao. Research on Classification Method of Micro Character of Paris Polyphylla with Efficient Attention Module[J]. Computer Engineering and Applications, 2022, 58(13): 272-279.

罗旭东, 李宗桂, 张俊华, 李学芳, 于文涛. 融合高效注意力的重楼微性状鉴别方法研究[J]. 计算机工程与应用, 2022, 58(13): 272-279.

References

[1] 王鑫，王艳翚.中药标准化问题与对策研究[J].中华中医药杂志，2018，33（1）：22-25.
WAN X，WANG Y H.Research on the standardization of Chinese medicine[J].China Journal of Traditional Chinese Medicine and Pharmacy，2018，33（1）：22-25.
[2] 符德欢，王丽，郭佳玉，等.4种重楼属药用植物的比较鉴别研究[J].云南中医学院学报，2017，40（4）：83-86.
FU D H，WANG L，GUO J Y，et al.Comparison study of 4 kinds of medicinal plants of the genus paris[J].Journal of Yunnan University of Traditional Chinese Medicine，2017，40（4）：83-86.
[3] 徐飞，孟沙，吴启南，等.基于卷积神经网络的人参与西洋参饮片鉴别方法研究[J].南京中医药大学学报，2018，34（6）：621-624.
XU F，MENG S，WU Q N，LOU Z，et al.Research on the differentiation of ginseng radix and panacis quinquefolii slices based on convolutional neural network[J].Journal of Nanjing University of Traditional Chinese Medicin，2018，34（6）：621-624.
[4] ANDAYANIU，SUMANTRI I B，PAHALA，et al.The implementation of deep learning using convolutional neural network to classify based on stomata microscopic image of curcuma herbal plants[C]//Proceedings of IOP Conference Series：Materials Science and Engineering，2020，851（1）：012035.
[5] 吴冲，谭超群，黄永亮，等.基于深度学习算法的川贝母、山楂及半夏饮片的智能鉴别[J].中国实验方剂学杂志，2020，26（21）：195-201.
WU C，TAN C Q，HUANG Y L，et al.Intelligent identification of fritillariae cirrhosae bulbus，crataegi fructus and pinelliae rhizoma based on deep learning algorithms[J].Chinese Journal of Experimental Traditional Medical Formulae，2020，26（21）：195-201.
[6] VISAVAKITCHAROEN A，RATANASANYA S，POLVICHA J.Improving thai herb image classification using convolutional neural networks with boost up features[C]//Proceedings of the 34th International Technical Conference on Circuits/Systems，Computers and Communications（ITC-CSCC），2019：1-4.
[7] 孙鑫，钱会南.基于深度卷积网络的中药饮片图像识别[J].世界科学技术-中医药现代化，2017，19（2）：218-222.
SUN X，QIAN H N.Identification of the images of Chinese herb slices with deep convolutional network[J].Modernization of Traditional Chinese Medicine and Materia Medica-World Science and Technology，2017，19（2）：218-222.
[8] KHRISNE D C，SUYADNYA I M A.Indonesian herbs and spices recognition using smaller VGGNet-like network[C]//Proceedings of the 2018 International Conference on Smart Green Technology in Electrical and Information Systems（ICSGTEIS），2018：221-224.
[9] LEE S H，CHAN C S，MAYO S J，et al.How deep learning extracts and learns leaf features for plant classification[J].Pattern Recognition，2017，71：1-13.
[10] QIN F，GAO N，PENG Y，et al.Fine-grained leukocyte classification with deep residual learning for microscopic images[J].Computer Methods and Programs in Biomedicine，2018，162：243-252.
[11] WANG B，WANG D.Plant leaves classification：A few-shot learning method based on siamese network[J].IEEE Access，2019，7：151754-151763.
[12] 孙丽萍，陈泓钢，岳琪，等.依据特征融合和深度学习的树木叶片分类方法[J].东北林业大学学报，2020，48（6）：51-55.
SUN L P，CHEN H G，YUE Q，et al.Classification methods of tree leaves by feature fusion and deep learning[J].Journal of Northeast Forestry University，2020，48（6）：51-55.
[13] VAN HIEU N，HIEN N L H.Automatic plant image identification of vietnamese species using deep learning models[J].International Journal of Emerging Trends & Technology in Computer Science，2020：2005.
[14] LIN Z，MU S，HUANG F，et al.A unified matrix-based convolutional neural network for fine-grained image classification of wheat leaf diseases[J].IEEE Access，2019，7：11570-11590.
[15] 王阳，刘立波.面向细粒度图像分类的双线性残差注意力网络[J].激光与光电子学进展，2020，57（12）：171-180.
WANG Y，LIU L B.Bilinear residual attention networks for fine-grained image classification[J].Laser & Optoelectronics Progress，2020，57（12）：171-180.
[16] ARAUJO V M，BRITTO A S，BRUN A L，et al.Fine-grained hierarchical classification of plant leaf images using fusion of deep models[C]//Proceeding of the IEEE 30th International Conference on Tools with Artificial Intelligence（ICTAI），2018：1-5.
[17] KAYA A，KECELI A S，CATAL C，et al.Analysis of transfer learning for deep neural network based plant classification models[J].Computers and Electronics in Agriculture，2019，158：20-29.
[18] NYASAKA D，WANG J，TINEGA H.Learning hyperspectral feature extraction and classification with ResNeXt network[J].arXiv：2002.02585，2020.
[19] ZHANG Q，ZHANG M，CHEN T，et al.Recent advances in convolutional neural network acceleration[J].Neurocomputing，2019，323：37-51.
[20] ZHENG H，FU J，MEI T，et al.Learning multi-attention convolutional neural network for fine-grained image recognition[C]//Proceedings of the IEEE International Conference on Computer Vision（ICCV），2017：5209-5217.
[21] ZHANG H，CISSE M，DAUPHIN Y N，et al.Mixup：Beyond empirical risk minimization[J].arXiv：1710. 09412，2017.
[22] SIMONYAN K，ZISSERMAN A.Very deep convolutional networks for large-scale image recognition[J].arXiv：1409.1556，2014.
[23] SZEGEDY C，IOFFE S，VANHOUCKE V，et al.Inception-v4，inception-resnet and the impact of residual connections on learning[C]//Proceedings of the AAAI Conference on Artificial Intelligence，2017.
[24] WANG Q，WU B，ZHU P，et al.ECA-Net：Efficient channel attention for deep convolutional neural networks[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition，2020：11534-11542.
[25] WOO S，PARK J，LEE J Y，et al.Cbam：Convolutional block attention module[C]//Proceedings of the European Conference on Computer Vision（ECCV），2018：3-19.
[26] HU J，SHEN L，SUN G.Squeeze-and-excitation networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2018：7132-7141.
[27] FENG J，LU S.Performance analysis of various activation functions in artificial neural networks[C]//Journal of Physics（Conference Series）.IOP Publishing，2019，1237（2）：022030.
[28] THARWAT A.Classification assessment methods[J/OL].Applied Computing and Informatics：1-13（2008）[2020-10-15].https：//doi.org/10.1016/j.aci.2018.08.003.2020.
[29] RUDER S.An overview of gradient descent optimization algorithms[J].arXiv：1609.04747，2016.