基于机器视觉的工夫红茶萎凋叶水分检测
|
摘要
萎凋是工夫红茶加工的首道工序,萎凋叶水分检测主要依赖于人工经验判断,难以做到精准、客观和量化评价。本文针对萎凋叶摊放工艺,以工夫红茶萎凋在制品为研究对象,基于机器视觉技术获取萎凋叶图像的色泽和纹理特征信息,分析图像特征变量的变化规律及其与水分的关联;采用偏最小二乘法PLS(Partial least squares)、极限学习机ELM(Exterme learning machine)和支持向量机回归SVR(Support vector machine regression)方法,分别建立萎凋叶水分定量预测模型。结果表明:3种预测方法预测值和实际值的相关系数均大于0.95,非线性SVR模型明显优于PLS模型和ELM模型,其预测集的相关系数Rp(Correlation coefficient of predication set)和预测均方根RMSEP(Root-mean-square error of prediction)分别为0.9904和0.0118,预测集绝对误差均小于0.05,相对标准偏差RPD(Relative percent deviation)值为6.6264。这说明SVR方法能够更好表征图像信息与水分之间的量化解析关系,具有更优的泛化性和鲁棒性,为萎凋加工中水分含量快速无损检测提供了解决方案,为工夫红茶水分在线检测装置的开发提供理论基础,对茶叶机械智能化的发展具有重要意义。
Withering is the first step in the processing of Congou black tea.The detection of the withered leaf moisture mainly depends on artificial experience,and it is difficult to achieve accurate,objective and quantitative evaluation.In the paper,for spreading process of withered leaf,taking the work in progress of black tea withering as the research object,based on machine vision technology,our obtained the color and texture feature information of the withering leaf images and analyzed the variation law of image feature variables and the relationship with moisture content.Using the methods of Partial least squares(PLS),Exterme learning machine(ELM) and Support vector machine regression(SVR) established a quantitative prediction model based the withering leaf moisture respectively.The results showed that the correlation coefficients between the predicted and actual values of the three prediction methods were all >0.95.The SVR model was better than the PLS model and the ELM model.The Correlation coefficient of predication set(Rp) values and the Root-mean-square error of calibration set(RMSEC) of the SVR model were 0.9904 and 0.0118,respectively.The absolute error of the prediction set were lower 0.05.The Relative percent deviation(RPD) value was 6.6264.It could be seen that Nonlinear modeling method of SVR could better characterize the quantitative analytical relationship between image information and moisture and had better generalization and robustness.The study provided a solution for the rapid non-destructive detecting of moisture content during the withering process,and a theoretical basis for Congfu black tea detection device of moisture online,which also produced a great significance to the development of tea machinery intelligence.
Withering is the first step in the processing of Congou black tea.The detection of the withered leaf moisture mainly depends on artificial experience,and it is difficult to achieve accurate,objective and quantitative evaluation.In the paper,for spreading process of withered leaf,taking the work in progress of black tea withering as the research object,based on machine vision technology,our obtained the color and texture feature information of the withering leaf images and analyzed the variation law of image feature variables and the relationship with moisture content.Using the methods of Partial least squares(PLS),Exterme learning machine(ELM) and Support vector machine regression(SVR) established a quantitative prediction model based the withering leaf moisture respectively.The results showed that the correlation coefficients between the predicted and actual values of the three prediction methods were all >0.95.The SVR model was better than the PLS model and the ELM model.The Correlation coefficient of predication set(Rp) values and the Root-mean-square error of calibration set(RMSEC) of the SVR model were 0.9904 and 0.0118,respectively.The absolute error of the prediction set were lower 0.05.The Relative percent deviation(RPD) value was 6.6264.It could be seen that Nonlinear modeling method of SVR could better characterize the quantitative analytical relationship between image information and moisture and had better generalization and robustness.The study provided a solution for the rapid non-destructive detecting of moisture content during the withering process,and a theoretical basis for Congfu black tea detection device of moisture online,which also produced a great significance to the development of tea machinery intelligence.
引文
[1]陈富桥.2017年国内茶叶产销形势通报[J].中国茶叶加工,2018(1):67.Chen F Q.2017 Domestic Tea Production and Marketing Situation Report[J].China Tea Processing,2018(1):67.
[2]韩余,肖宏儒,秦广明,等.红茶加工工艺及机械设备研究进展[J].中国农机化学报,2013,34(2):20-25.Han Y,Xiao H R,Qin G M,et al.Current reserch situations about processing technic and mechanic equipments for black tea[J].Journal of Chinese Agricultural Mechanization,2013,34(2):20-25
[3]李文萃,任广鑫,何鑫,等.茶鲜叶摊放与萎凋技术及设备应用现状与改进[J].中国茶叶加工,2013(3):44-48.Li W C,Ren G X,He X,et al.Status and Improvements of the Technology and Equipment in Tea-leaf Spreading and Withering[J].China Tea Processing,2013(3):44-48.
[4]陈寿松,占杨,郑功宇,等.茶叶含水率常用测定方法及比较分析[J].中国茶叶加工,2013(3):33-36.Chen S S,Zhan Y,Zhen G Y,et al.Common Determination Methods and Comparative Analysis for Moisture Content of Tea[J].China Tea Processing,2013(3):33-36.
[5]方观玉,郑文慧,卢秋华,等.3种茶叶中水分含量检测方法比较[J].福建茶叶,2014,36(4):20-22.Fang G Y,Zheng W H,Lu Q H,et al.Comparison of water content detection methods in three kinds of teas[J].Fujian Tea,2014,36(4):20-22.
[6]黄藩,董春旺,朱宏凯,等.红茶萎凋中鲜叶理化变化及工艺技术研究进展[J].中国农学通报,2014,30(27):275-281.Huang F,Dong C W,Zhu H K,et al.A Review on the Black Tea Withering Technology and the Changes of Physics and Chemistry of Leaves[J].Chinese Agricultural Science Bulletin,2014,30(27):275-281.
[7]李洁,齐桂年.利用计算机读取整茶色泽参数的方法研究[J].茶叶科学,2007(4):328-334.Li J,Qi G N.Research on the Method of Reading the Color Parameters of Whole Tea by Computer Journal of Tea Science,2007(4):328-334.
[8]吴正敏,曹成茂,谢承健,等.基于图像处理技术和神经网络实现机采茶分级[J].茶叶科学,2017,37(2):182-190.Wu Z M,Cao C M,Xie C J,et al.Grading of Machine Picked Tea Based on Image Processing Technology and Neural Network[J].Journal of Tea Science,2017,37(2):182-190.
[9]王财盛,朱威,徐召飞,等.基于机器视觉的采茶机割刀控制方法[J].计算机测量与控制,2017,25(4):70-74.Wang C S,Zhu W,Xu Z F,et al.Cutter Control Method of Tea-picking Machine Based on Machine Vision[J].Computer Measurement&Control,2017,25(4):70-74.
[10]Zhang Y D,Yang X J,Carlo Cattani,et al.Tea category identification using a novel fractional fourier entropy and jaya algorithm[J].Entropy,2016,18(3):77.
[11]Wang S H,Yang X J,Zhang Y D,et al.Identification of green,oolong and black teas in china via wavelet packet entropy and fuzzy support vector machine[J].Entropy,2015,17(10):6663-6682.
[12]贾广松.基于茶鲜叶表面可见特征与含水率变化模型研究[D].杭州:浙江工业大学,2016.
[13]牛亚晓,张立元,韩文霆.基于Lab颜色空间的棉花覆盖度提取方法研究[J].农业机械学报,2018,49(10):240-249.Niu Y X,Zhang L Y,Han W T.Extraction Methods of Cotton Coverage Based on Lab Color Space[J].Transactions of the Chinese Society for Agricultural Machinery,2018,49(10):240-249.
[14]吴迪,刘伟峰,胡胜,等.基于Lab空间的K均值聚类彩色图像分割[J].电子科技,2017,30(10):29-32.Wu D,Liu W F,Hu S,et al.Color Image Segmentation Using Kmean Clustering Based on Lab Space[J].Electronic Science&Technology,2017,30(10):29-32.
[15]Liang G Z,Dong C W,Hu B,et al.Prediction of moisture content for congou black tea withering leaves using image features and nonlinear method[J].Scientific reports,2018,8(1):7854.
[16]Dong C W,Zhu H K,Zhao J W,et al.Sensory quality evaluation for appearance of needle-shaped green tea based on computer vision and nonlinear tools[J].J Zhejiang Univ Sci B,2017,18(6):544-548.
[17]袁剑波,毛红日.基于马氏距离的数据处理方法及其在高速公路改扩建工程定额中的应用[J].铁道科学与工程学报,2018,15(10):2715-2720.Yuan J B,Mao H R.The data processing method based on Mahalanobis distance and its application inhighway reconstruction engineering quota[J].Journal of Railway Science and Engineering,2018,15(10):2715-2720.
[18]李丹,陈岗,练学燕,等.自动化生产线加工过程中“川红芽茶”主要品质成分变化及回归分析[J].食品工业科技,2015,36(23):79-83,88.Li D,Chen G,Lian X Y,et al.Variation and regression analysis in automated production lines of Sichuan black bud tea[J].Science&Technology of Food Industry,2015,36(23):79-83,88.
[19]范珊珊,郭海朋,朱菊艳,等.线性回归模型在北京平原地面沉降预测中的应用[J].中国地质灾害与防治学报,2013,24(1):70-74.Fan S S,Guo H P,Zhu J Y,et al.Application of linear regression model for land subsidence prediction in Beijing plain[J].Chinese Journal of Geological Hazard&Control,2013,24(1):70-74.
[20]袁海波,滑金杰,邓余良,等.不同环境湿度条件下萎凋叶色差变化分析[J].茶叶科学,2015,35(6):551-558.Yuan H B,Hua J J,Deng Y L,et al.Analysis on the Changes of Color of Withering Leaf in Different Environmental Humidity Conditions[J].Journal of Tea Science,2015,35(6):551-558.
[21]黄藩,董春旺,高明珠,等.工夫红茶萎凋中温度对鲜叶失水率影响的预测模型[J].中国农学通报,2014,30(34):193-198.Huang F,Dong C W,Gao M Z,et al.Predictive Mathematical Model of Temperature on Water Losing Rate of Leaves in Withering Process of Congou Black Tea[J].Chinese Agricultural Science Bulletin,2014,30(34):193-198.
[22]钟江,单保淋,乔欣.基于微流体水分散失的茶叶杀青研究[J].轻工机械,2016,34(3):14-17.Zhong J,Shan B L,Qiao X.Research of Tea Fixation Based on Microfluid[J].Light Industry Machinery,2016,34(3):14-17.
[2]韩余,肖宏儒,秦广明,等.红茶加工工艺及机械设备研究进展[J].中国农机化学报,2013,34(2):20-25.Han Y,Xiao H R,Qin G M,et al.Current reserch situations about processing technic and mechanic equipments for black tea[J].Journal of Chinese Agricultural Mechanization,2013,34(2):20-25
[3]李文萃,任广鑫,何鑫,等.茶鲜叶摊放与萎凋技术及设备应用现状与改进[J].中国茶叶加工,2013(3):44-48.Li W C,Ren G X,He X,et al.Status and Improvements of the Technology and Equipment in Tea-leaf Spreading and Withering[J].China Tea Processing,2013(3):44-48.
[4]陈寿松,占杨,郑功宇,等.茶叶含水率常用测定方法及比较分析[J].中国茶叶加工,2013(3):33-36.Chen S S,Zhan Y,Zhen G Y,et al.Common Determination Methods and Comparative Analysis for Moisture Content of Tea[J].China Tea Processing,2013(3):33-36.
[5]方观玉,郑文慧,卢秋华,等.3种茶叶中水分含量检测方法比较[J].福建茶叶,2014,36(4):20-22.Fang G Y,Zheng W H,Lu Q H,et al.Comparison of water content detection methods in three kinds of teas[J].Fujian Tea,2014,36(4):20-22.
[6]黄藩,董春旺,朱宏凯,等.红茶萎凋中鲜叶理化变化及工艺技术研究进展[J].中国农学通报,2014,30(27):275-281.Huang F,Dong C W,Zhu H K,et al.A Review on the Black Tea Withering Technology and the Changes of Physics and Chemistry of Leaves[J].Chinese Agricultural Science Bulletin,2014,30(27):275-281.
[7]李洁,齐桂年.利用计算机读取整茶色泽参数的方法研究[J].茶叶科学,2007(4):328-334.Li J,Qi G N.Research on the Method of Reading the Color Parameters of Whole Tea by Computer Journal of Tea Science,2007(4):328-334.
[8]吴正敏,曹成茂,谢承健,等.基于图像处理技术和神经网络实现机采茶分级[J].茶叶科学,2017,37(2):182-190.Wu Z M,Cao C M,Xie C J,et al.Grading of Machine Picked Tea Based on Image Processing Technology and Neural Network[J].Journal of Tea Science,2017,37(2):182-190.
[9]王财盛,朱威,徐召飞,等.基于机器视觉的采茶机割刀控制方法[J].计算机测量与控制,2017,25(4):70-74.Wang C S,Zhu W,Xu Z F,et al.Cutter Control Method of Tea-picking Machine Based on Machine Vision[J].Computer Measurement&Control,2017,25(4):70-74.
[10]Zhang Y D,Yang X J,Carlo Cattani,et al.Tea category identification using a novel fractional fourier entropy and jaya algorithm[J].Entropy,2016,18(3):77.
[11]Wang S H,Yang X J,Zhang Y D,et al.Identification of green,oolong and black teas in china via wavelet packet entropy and fuzzy support vector machine[J].Entropy,2015,17(10):6663-6682.
[12]贾广松.基于茶鲜叶表面可见特征与含水率变化模型研究[D].杭州:浙江工业大学,2016.
[13]牛亚晓,张立元,韩文霆.基于Lab颜色空间的棉花覆盖度提取方法研究[J].农业机械学报,2018,49(10):240-249.Niu Y X,Zhang L Y,Han W T.Extraction Methods of Cotton Coverage Based on Lab Color Space[J].Transactions of the Chinese Society for Agricultural Machinery,2018,49(10):240-249.
[14]吴迪,刘伟峰,胡胜,等.基于Lab空间的K均值聚类彩色图像分割[J].电子科技,2017,30(10):29-32.Wu D,Liu W F,Hu S,et al.Color Image Segmentation Using Kmean Clustering Based on Lab Space[J].Electronic Science&Technology,2017,30(10):29-32.
[15]Liang G Z,Dong C W,Hu B,et al.Prediction of moisture content for congou black tea withering leaves using image features and nonlinear method[J].Scientific reports,2018,8(1):7854.
[16]Dong C W,Zhu H K,Zhao J W,et al.Sensory quality evaluation for appearance of needle-shaped green tea based on computer vision and nonlinear tools[J].J Zhejiang Univ Sci B,2017,18(6):544-548.
[17]袁剑波,毛红日.基于马氏距离的数据处理方法及其在高速公路改扩建工程定额中的应用[J].铁道科学与工程学报,2018,15(10):2715-2720.Yuan J B,Mao H R.The data processing method based on Mahalanobis distance and its application inhighway reconstruction engineering quota[J].Journal of Railway Science and Engineering,2018,15(10):2715-2720.
[18]李丹,陈岗,练学燕,等.自动化生产线加工过程中“川红芽茶”主要品质成分变化及回归分析[J].食品工业科技,2015,36(23):79-83,88.Li D,Chen G,Lian X Y,et al.Variation and regression analysis in automated production lines of Sichuan black bud tea[J].Science&Technology of Food Industry,2015,36(23):79-83,88.
[19]范珊珊,郭海朋,朱菊艳,等.线性回归模型在北京平原地面沉降预测中的应用[J].中国地质灾害与防治学报,2013,24(1):70-74.Fan S S,Guo H P,Zhu J Y,et al.Application of linear regression model for land subsidence prediction in Beijing plain[J].Chinese Journal of Geological Hazard&Control,2013,24(1):70-74.
[20]袁海波,滑金杰,邓余良,等.不同环境湿度条件下萎凋叶色差变化分析[J].茶叶科学,2015,35(6):551-558.Yuan H B,Hua J J,Deng Y L,et al.Analysis on the Changes of Color of Withering Leaf in Different Environmental Humidity Conditions[J].Journal of Tea Science,2015,35(6):551-558.
[21]黄藩,董春旺,高明珠,等.工夫红茶萎凋中温度对鲜叶失水率影响的预测模型[J].中国农学通报,2014,30(34):193-198.Huang F,Dong C W,Gao M Z,et al.Predictive Mathematical Model of Temperature on Water Losing Rate of Leaves in Withering Process of Congou Black Tea[J].Chinese Agricultural Science Bulletin,2014,30(34):193-198.
[22]钟江,单保淋,乔欣.基于微流体水分散失的茶叶杀青研究[J].轻工机械,2016,34(3):14-17.Zhong J,Shan B L,Qiao X.Research of Tea Fixation Based on Microfluid[J].Light Industry Machinery,2016,34(3):14-17.