摘要
为提升烟叶含青程度识别效率和准确度,通过对不同含青程度烟叶参比样的构建,采用计算机图像处理技术筛选出含青区域提取的最优颜色通道和参数阈值,并基于支持向量机(SVM)构建不同含青程度烟叶的识别模型.结果表明:烟叶正面透光图像适合用于含青区域的提取;Lab颜色模型的a通道提取图像中含青区域的效果最佳,其阈值范围为(141,142);SVM模型对不同程度含青烟叶样品的测试准确率达到86.01%.混淆矩阵显示,SVM模型对正组和微带青烟叶的识别准确率较高.图像处理技术和支持向量机对含青烟叶程度的划分有较好的应用效果.
In order to improve the efficiency and accuracy of containing cyan degree identification of tobacco leaves.By constructing tobacco leaf reference samples with different containing cyan content.The best color channel and parameter threshold for the extraction of cyan region are screened by computer image processing technology.Construction of recognition model of tobacco leaves with different green degree based on support vector machine.The front transparent image of tobacco leaf is suitable for the extraction of containing cyan areas.The a-channel of Lab color model is the best way to extract the cyan region in the image.And its threshold range is(141,142).The test accuracy of SVM model for different green tobacco samples reached 86.01%.The confusion matrix shows that the recognition accuracy of the model for positive group and microstrip green tobacco leaves is superior.Image processing technology and support vector machine have good application effect on the division of containing cyan tobacco leaf.
作者
李峥
徐志强
张晓兵
林珈夷
钟永健
徐均华
张赵鹏
焦得平
LI Zheng;XU Zhiqiang;ZHANG Xiaobing;LIN Jiayi;ZHONG Yongjian;XU Junhua;ZHANG Zhaopeng;JIAO Deping(Technology Center of China Tobacco Zhejiang Industrial Co.,Ltd.,Hangzhou,Zhejiang,China 310024;Shanghai Micro Vision Technology LTD,Shanghai,China 200082)
出处
《昆明学院学报》
2023年第6期19-25,共7页
Journal of Kunming University
基金
浙江中烟工业有限责任公司重点科技项目(ZJZY2021B001).
关键词
初烤烟叶
含青程度
支持向量机
图像处理
识别模型
flue-cured tobacco
degree of containing cyan
support vector machine
image processing
identification model
作者简介
李峥,男,浙江杭州人,浙江中烟工业有限责任公司助理农艺师,硕士,研究方向为烟叶外观质量数字化评价;通信作者:徐志强,男,山东莒县人,浙江中烟工业有限责任公司烟叶分级高级技师,研究方向为烟叶质量评价与工业应用,E-mail:xuzq@zjtobacco.com.
