北京生物醫(yī)學(xué)工程

基于深度學(xué)習(xí)的中藥材飲片圖像識別

Identification of the images of Chinese herb slices with deep learning

作者：劉加峰高子嘯段元民李海云石宏理

單位：首都醫(yī)科大學(xué)生物醫(yī)學(xué)工程學(xué)院（北京100069）通信作者:石宏理，副教授。E-mail: [email protected] <p> </p>

關(guān)鍵詞：中藥飲片；圖像識別;深度學(xué)習(xí);卷積神經(jīng)網(wǎng)絡(luò);學(xué)習(xí)算法

分類號：R318. 04 <p> </p>

出版年·卷·期（頁碼）：2021·40·6（605-608）

摘要：

目的建立一個基于深度卷積神經(jīng)網(wǎng)絡(luò)的中藥飲片圖像檢測識別系統(tǒng)。該系統(tǒng)對于正常情況下采集的中藥飲片圖像，能夠自動檢測識別出相應(yīng)類別的中藥飲片。方法本文使用了 SSD目標(biāo)檢測算法，構(gòu)建數(shù)據(jù)集，利用標(biāo)注工具進(jìn)行了標(biāo)注，然后在云端colab上進(jìn)行調(diào)試代碼、訓(xùn)練、測試、驗(yàn)證。結(jié)果對于3種中藥飲片（枸杞、甘草、陳皮）進(jìn)行識別驗(yàn)證，平均識別率高于80%,樣本集足夠大可以有效提高識別準(zhǔn)確率。結(jié)論本文將卷積神經(jīng)網(wǎng)絡(luò)應(yīng)用于中藥材識別中，將傳統(tǒng)的中醫(yī)學(xué)與新興的深度學(xué) 習(xí)網(wǎng)絡(luò)相結(jié)合,識別中藥飲片的效率高,速度快，準(zhǔn)確率高，可應(yīng)用于絕大部分需要識別中藥飲片類別的場景。

Objective A deep learning-convolutional neural network based image detection and recognition system for Chinese herbal slices is built. The system is capable of automatically detecting and recognizing categories and locating the images of traditional Chinese medicine drinking tablets that contain multiple categories established under simulated normal conditions. Methods In this paper, we used the SSD target detection algorithm ? established the image database, labeled them using the labeling tool, and then debugged the code, trained, tested, and verified them on the cloud colab. Results For the three Chinese herbal slices ( Chinese wolfberry, licorice, and pericarpium citri reticulatae) , the average recognition rate was more than 80%, and in particular, if the sample set was large enough then the recognition accuracy was improved. Conclusions In this paper, convolution neural network is applied to the identification of traditional Chinese herbs, which combines traditional Chinese medicine with the new deep learning network. It has high efficiency, fast speed and high accuracy. It can be applied to most scenes that need to identify the categories of traditional Chinese herbs.

參考文獻(xiàn)：

[1 ]孫鑫,錢會南.基于深度卷積網(wǎng)絡(luò)的中藥飲片圖像識別[J] .世界科學(xué)技術(shù)一中醫(yī)藥現(xiàn)代化,2017,19(2) :218-222.

Sun X, Qian HN. 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.

[2 ]陶歐,張燕玲，陳茜，等.基于灰度共生矩陣的中藥飲片橫切面圖像紋理特征參數(shù)的提取[J].世界科學(xué)技術(shù)一中醫(yī)藥現(xiàn)代化,2014,16(12) :2531-2537.

Tao 0, Zhang YL, Chen Q, et aL Extraction of texture feature parameter of transverse section in Chinese herbal medicine by gray-level co-occurrence matrix[ J]. Modernization of Traditional Chinese Medicine and Materia Medica-World Science and Technology,2014,16( 12) :2531-2537.

[3]陶歐，林兆洲,張憲寶，等.基于飲片切面圖像紋理特征參數(shù)的中藥辨識模型研究[J].世界科學(xué)技術(shù)一中醫(yī)藥現(xiàn)代化,2014,16(12)：2558-2562.Tao 0, Lin ZZ, Zhang XB, et al. Research on identification model of Chinese herbal medicine by texture feature parameter of transverse section image [ J ]. Modernization of Traditional Chinese Medicine and Materia Medica-World Science and Technology, 2014,16(12):2558-2562.

[4]夏黎明，沈堅(jiān)，張榮國，等.深度學(xué)習(xí)技術(shù)在醫(yī)學(xué)影像領(lǐng)域的應(yīng) 用[J].協(xié)和醫(yī)學(xué)雜志,2018,9(1) :10-14.

Xia LM, Shen J, Zhang RG, et al. Application of deep learning in medical imaging research [ J]. Medical Journal of Peking Union Medical College Hospital,2018,9( 1):10-14.

[5 ]張榮磊，田愛奎，譚浩，等.基于卷積神經(jīng)網(wǎng)絡(luò)的圖像識別算法研究[J].山東理工大學(xué)學(xué)報(自然科學(xué)版),2018,32( 1):48-51.

Zhang RL,Tian AK,Tan H,et al. Research on image recognition algorithm based on convolutional neural networks [ J]. Journal of Shandong University of Technology ( Natural Science Edition ), 2018,32(1):48-51.

[6 ]林明旺.基于卷積神經(jīng)網(wǎng)絡(luò)的魚類圖像識別與分類[J].電子技術(shù)與軟件工程,2017(6) :92-93.

[7]秦東輝，周輝，趙雄波，等.基于卷積神經(jīng)網(wǎng)絡(luò)圖像識別算法的加速實(shí)現(xiàn)方法[J].航天控制,2019,37(1):21-26.

Qin DH, Zhou H, Zhao XB, et al. Accelerated implementation method of image recognition algorithm based on convolutional neural network[ J]. Aerospace Control,2019,37(1) :1-26.

[8 ] Liu JX,Xu BL, Shen LL, et al. HEp-2 cell classification based on a deep autoencoding-classification convolutional neural network [ C ] // 2017 IEEE 14th International Symposium on Biomedical Imaging. Melbourne, VIC , Australia: IEEE Press, 2017:1019-1023.

[9 ] Windrim L, Melkumyan A, Murphy RJ, et al. Pre-training for hyperspectral convolutional neural network classification [ J ]. IEEE Transactions on Geoscience & Remote Sensing, 2018,56 (5) ：2798-2810.

[10]劉偉，陳鴻昶,黃瑞陽.基于Tree-based CNN的關(guān)系抽取[J]. 中文信息學(xué)報,2018,32( 11) :34-40.

Liu W, Chen HC , Huang RY. Neural relation extraction via treebased CNN[ J]. Journal of Chinese Information Processing,2018, 32( 11):34-40.

[11 ] Jonghong Kim, Gil-Jin Jang, Minho Lee. Investigation of the efficiency of unsupervised learning for multi-task classification in convolutional neural network [ M ]//International Conference on Neural Information Processing. ICONIP 2016. Lecture Notes in Computer Science. Switzerland: Springer, Cham, 2016, 9949:547-554

服務(wù)與反饋：

【文章下載】【加入收藏】

提示：您還未登錄，請登錄！點(diǎn)此登錄

51黑料吃瓜在线观看,51黑料官网|51黑料捷克街头搭讪_51黑料入口最新视频