基于图像识别的烟丝结构检测及烟丝组分分析
Cut tobacco structure detection and cut tobacco component analysis based on image recognition
-
摘要: 为提高烟丝结构检测及烟丝组分分析的准确性,基于图像识别技术,对烟丝轮廓进行细化,提取烟丝骨骼,得到烟丝长度,建立烟丝表观总面积与质量的拟合模型,获得烟丝结构(整丝率、碎丝率);利用最小内切圆的方法得到烟丝的平均宽度、宽度方差,烟丝轮廓在饱和度(Saturation,S)通道上的颜色方差,以及HSV颜色模型的颜色矩,用支持向量机(Support Vector Machine,SVM)作为分类器,构建薄片丝、梗丝、叶丝的烟丝组分分类模型。实际应用结果表明:基于图像识别的方法能准确统计整丝率、碎丝率,且比传统振筛法更快捷、有效,与卷积神经网络法和残差神经网络法相比,该方法识别薄片丝、梗丝、叶丝的平均相对误差≤5%,准确性及可行性更高。Abstract: In order to improve the accuracy of cut tobacco structure detection and cut tobacco component analysis, based on image recognition technology to thinning the contour of cut tobacco image, extract the skeleton of cut tobacco, get the length of cut tobacco, and establish a fitting model for the apparent total area and mass of cut tobacco and get cut tobacco structure (whole cut rate and broken cut rate). The method of the smallest inscribed circle was used to obtain the average width and width variance of the cut tobacco, the color variance of the cut tobacco profile on the Saturation (S) channel, and the color moment of the HSV color model, the Support Vector Machine (SVM) was used as the classifier to construct the tobacco component classification model of tobacco flakes, cut stem, and cut tobacco. The practical application results showed that the method based on image recognition could accurately count whole cut rate and broken cut rate and was faster and more effective than the quality-control shake method. Compared with the convolutional neural network method and the residual neural network method, the average relative error of the method for identifying tobacco flakes, cut stem, and cut tobacco was less than 5%, and the accuracy and feasibility were higher.
-
Key words:
- tobacco structure /
- tobacco component /
- image recognition /
- SVM /
- HSV
-
-
[1]
唐莹.打叶复烤工艺参数优化讨论[J].科技传播,2013(16):134-135.
-
[2]
罗海燕,方文青,杨林波,等.叶中含梗率与相关打叶质量指标的关系[J].烟草科技,2005(7):11-13.
-
[3]
申晓锋,李华杰,李善莲,等.烟丝结构表征方法研究[J].中国烟草学报,2010,16(2):20-25.
-
[4]
楚晗,范磊,王爱霞,等.不同规格卷烟烟丝结构与物理质量的差异性研究[J].南方农业学报,2019, 50(7):1565-1571.
-
[5]
郭三刺,申超群.基于图像处理技术的烟丝检测系统[J].河南科技,2014(11):21-22.
-
[6]
钟宇,周明珠,徐燕,等.基于残差神经网络的烟丝类型识别方法的建立[J].烟草科技,2021,54(5):82-89.
-
[7]
单凯.烟丝整丝率影响因素问题探讨[J].中国科技博览,2013(38):397.
-
[8]
国家烟草专卖局.卷烟工艺规范[M].北京:中国轻工业出版社,2017:174.
-
[9]
CHEN J S,KANG X G,YE L,et al.Median filtering forensics based on convolutional neural networks[J].IEEE Signal Processing Letters,2015,22(11):1849-1853.
-
[10]
WANG R,LI C,WANG J,et al.Threshold segmentation algorithm for automatic extraction of cerebral vessels from brain magnetic resonance angiography images[J].Journal of Neuroscience Methods,2015(241):30-36.
-
[11]
张晋阳,孙樊珩.手背静脉图像骨架特征提取的算法[J].计算机应用,2007(1):152-154.
-
[12]
张敏.二值图像中一种改进的细化算法[J].现代电子技术,2005(19):96-98.
-
[13]
李善莲,申晓锋,李华杰,等.烟丝结构对卷烟端部落丝量的影响[J].烟草科技,2010(2):5-7,10.
-
[14]
李继英,王彬彬.降低薄片丝结团率的方法探析[J].中国科技信息,2013(6):99.
-
[15]
关丽娜.梗丝结构对卷烟质量稳定性的影响[J].中国科技博览,2013(35):343.
-
[16]
胡建军,周冀衡,熊燕,等.烘丝工艺参数对烘后叶丝质量影响的研究[J].中国烟草学报,2007(6):24-29.
-
[17]
田庆国,葛宝臻,李云鹏,等.利用轮廓线多边形表示实时提取光带中心线[J].光学精密工程,2012,20(6):1357-1364.
-
[18]
付忠良.图象阈值选取方法的构造[J].中国图象图形学报,2000(6):21-24.
-
[19]
魏赟,欧阳鹏.基于快速亮通双边滤波器的Retinex图像增强算法[J].小型微型计算机系统,2021,42(9):1944-1949.
-
[20]
龙浩,李庆党,张明月.基于HSV颜色空间和局部纹理的阴影消除算法研究[J].电子测量技术,2020,43(18):81-87.
-
[21]
国家烟草专卖局.烟草整丝率、碎丝率的测定方法:YC/T 178-2003[S].北京:中国标准出版社,2003.
-
[22]
高震宇,王安,董浩,等.基于卷积神经网络的烟丝物质组成识别方法[J].烟草科技,2017,50(9):68-75.
-
[1]
-
点击查看大图
计量
- PDF下载量: 108
- 文章访问数: 5192
- 引证文献数: 0
下载:
