本文已被:浏览 1466次 下载 995次 |
![本文二维码信息](/download_upload_file.aspx?file_name=/uploadfile/two_dimensional_code_img/zgnydxxb/2022/5/20220511.png) 码上扫一扫! |
基于机器视觉技术的小粒中药材种子净度快速检测 |
程莹1,许亚男1,侯浩楠1,宁翠玲2,杨成民3,董学会1,曹海4*,孙群1*
|
|
(1.中国农业大学 农学院/农业农村部农作物种子全程技术研究北京创新中心/北京市作物遗传改良重点实验室, 北京 100193;2.承德恒德本草农业科技有限公司, 河北 承德067000;3.中国医学科学院 药用植物研究所, 北京 100193;4.恒德本草(北京)农业科技有限公司, 北京 100070) |
|
摘要: |
为探究机器视觉技术用于小粒中药材种子净度快速检测的可行性,以黄芩、桔梗、黄芪、紫苏和柴胡5种常见小粒中药材种子为材料,使用扫描仪获取净种子、其他植物种子和所含杂质的图像,采用种子自动化分析系统(PhenoSeed)批量提取种子、其他植物种子及所含杂质的颜色、尺寸及纹理信息,通过相关性分析和主成分分析进行特征变量的筛选,采用多层感知器(MLP)和二元逻辑回归(BLR)建立上述5种中药材种子净度快速检测模型。结果表明,净种子、其他植物种子及所含杂质在物理指标方面存在显著差异,针对不同种子,采用不同指标建立的MLP净度模型的训练集和测试集准确率均在96.0%以上,该模型在不同中药材种子上的稳定性均优于BLR模型;以特征指标建立的模型稳定性优于全部指标的建模效果,运用特征变量建立的MLP模型对不同净度梯度(75.0%~100.0%)的混合样本进行预测,回归曲线的决定系数均达到0.99以上。采用机器视觉技术获取种子、其他植物种子及所含杂质颜色、尺寸和纹理等信息,以特征指标建立MLP模型可用于小粒中药材种子的净度快速检测。 |
关键词: 小粒中药材种子 净度 机器视觉 多层感知器(MLP) 二元逻辑回归(BLR) |
DOI:10.11841/j.issn.1007-4333.2022.05.11 |
投稿时间:2021-06-10 |
基金项目:国家中医药管理局“科技助力经济2020”重点专项(202004610111024);大同市与中国农业大学市校合作项目(201904710111639) |
|
Rapid small seed clarity detection method of Chinese medicinal plants based on machine vision technology |
CHENG Ying1,XU Yanan1,HOU Haonan1,NING Cuiling2,YANG Chengmin3,DONG Xuehui1,CAO Hailu4*,SUN Qun1*
|
(1.College of Agronomy and Biotechnology/The Innovation Center(Beijing)of Crop Seeds Whole-Process TechnologyResearch of Ministry of Agriculture and Rural Affairs/Beijing Key Laboratory of Crop Genetic Improvement, China Agricultural University, Beijing 100193, China;2.Chengde Hengde Materia Medica Agricultural Technology Co., Ltd., Chengde 067000, China;3.The Institute of Medicinal Plant Development, Chinese Academy of Medical Sciences, Beijing 100193, China;4.Hengde Materia Medica(Beijing)Agricultural Technology Co., Ltd., Beijing 100070, China) |
Abstract: |
To explore the applicability of machine vision technology in the rapid clarity detection of the seeds of Chinese medicinal plants, five kinds of common small seeds of Chinese medicinal plants, namely Scutellaria baicalensis Georgi, Platycodon grandiflorum(Jacq. )A. DC. , Astragalus membranaceus(Fisch. )Bge. , Perilla frutescens(L. )Britt. , and Bupleulum chinense DC. were collected in this study. A scanner was used to obtain the images of purity seeds, seeds of other plants and impurities. The automatic seed analysis system(PhenoSeed)was then used for batch extraction of the color, size and texture information of purity seeds, seeds of other plants and impurities. Characteristic variables were obtained by correlation analysis and principal component analysis, and multilayer perceptron(MLP)and binary logistic regression(BLR)were adopted to construct rapid seed clarity detection model for the five kinds of Chinese medicinal plants. The results showed that: There were significant differences in the physical features among purity seeds, seeds of other plants and impurities. The accuracies of both the training set and testing set of the MLP clarity model based on different features of different seeds were above 96. 0%. The stability of the MLP model performed better than that of the BLR model for different seeds. Moreover, it was found that the stability of model by characteristic features was better than the model effect of all 54 features. The characteristic features of MLP model was verified with mixed samples with different clarity gradients(75. 0%-100. 0%), and the determination coefficients of regression curves were all above 0. 99. Machine vision technology was used to obtain the color, size and texture information of purity seeds, seeds of other plants and impurities. In conclusion, the MLP model based on above characteristic features can be used to detect the clarity of the seeds of small Chinese medicinal plants quickly. |
Key words: small seeds of Chinese medicinal plants clarity machine vision multilayer perceptron(MLP) binary logistic regression(BLR) |