基于主成分分析的农药在线混合均匀性分析方法
|
摘要
农药在线混合是农药精确使用的有效途径,其中在线混合均匀性是影响施药效果的关键因素。利用高速摄像和图像处理技术结合主成分分析方法(PCA),研究了农药在线混合均匀性的计算方法。在混药浓度一定的条件下,研究了不同流量下的混药流场,利用高速摄像系统采集图像:提取兴趣区域图像(100×300)并进行均值滤波和简单缩放预处理,选择感受野大小为50×50像素、步距为5像素的步距式搜索方式来选取感受野子图,并对感受野子图进行尺寸为2×2像素、步距为2像素的均值池化;设置累积方差贡献率等于98%,对所有感受野子图进行PCA分析,可视化主成分特征和主成分系数分布,根据主成分特征相似性来可视化均匀性和定量计算不均匀性值。其中,当流量逐渐减小时,农药在线混合不均匀性值均值逐渐增大,均匀性变差。研究表明,利用高速摄像技术和主成分分析方法计算农药在线混合均匀性是可行的,为农药在线混合均匀性评价提供了一个新途径。
Pesticide spraying is an effective way to prevent from pest attacks and increase agricultural crop yields.However,current pesticide application methods and equipment are inefficient,resulting in wasting pesticides,environmental pollution,and serious ill-health even death to human and animals. On-line-mixing inside the pesticide mixers is an effective way to use pesticides precisely and the on-line-mixing uniformity is the key factor that affects the pesticide application. Therefore,howto measure the uniformity of pesticide on-line-mixing is needed to be solved currently. In this study,the high-speed imaging technology and principal component analysis(PCA) were used to analyze the uniformity of pesticide on-line-mixing as the brightness difference between water and pesticides with fluorescent substance,which could be distinguished easily under the ultraviolet lamp. Under the condition of a certain mixing concentration,the stream fields under eight different flowrates were used as study objects and images were collected by using an image acquisition system. Firstly,the region of interest(ROI) of 20 images in every flowrate were acquired(100×300),and processed by mean filtering and simple rescaling from the range [0,255] to [0,1]. The receptive fields were selected by the way of stride search with a size of 50× 50 pixels and stride of 5 pixels,and processed by the average pooling with a size of 2×2 pixels and stride of 2 pixels. Secondly,the statistical method of principal component analysis with 98% percentage of variance retained was used to establish a generative model for all receptive fields,and the principal component features and the principal component coefficients distribution were visualized for easy-understanding the PCA results of pesticide on-line-mixing images. The visualization showed a little clustering effect of receptive fields in eight different flowrates. Finally,the uniformity was visualized through the similarity heap maps and the non-uniformity value was calculated according to the feature similarity of the principal components. The experimental results showed that,when the flowrate decreased gradually,the average non-uniformity value increased from 6.492 to 18.262 and the standard deviation increased from 0.536 to 3.625,and the uniformity of pesticide on-line-mixing became worse. The results revealed that the proposed method is a promising method to evaluate uniformity of pesticide on-line-mixing. This method provided powerful technical support for the optimization of the structure of mixers and for the realization of the precision pesticide spraying.
Pesticide spraying is an effective way to prevent from pest attacks and increase agricultural crop yields.However,current pesticide application methods and equipment are inefficient,resulting in wasting pesticides,environmental pollution,and serious ill-health even death to human and animals. On-line-mixing inside the pesticide mixers is an effective way to use pesticides precisely and the on-line-mixing uniformity is the key factor that affects the pesticide application. Therefore,howto measure the uniformity of pesticide on-line-mixing is needed to be solved currently. In this study,the high-speed imaging technology and principal component analysis(PCA) were used to analyze the uniformity of pesticide on-line-mixing as the brightness difference between water and pesticides with fluorescent substance,which could be distinguished easily under the ultraviolet lamp. Under the condition of a certain mixing concentration,the stream fields under eight different flowrates were used as study objects and images were collected by using an image acquisition system. Firstly,the region of interest(ROI) of 20 images in every flowrate were acquired(100×300),and processed by mean filtering and simple rescaling from the range [0,255] to [0,1]. The receptive fields were selected by the way of stride search with a size of 50× 50 pixels and stride of 5 pixels,and processed by the average pooling with a size of 2×2 pixels and stride of 2 pixels. Secondly,the statistical method of principal component analysis with 98% percentage of variance retained was used to establish a generative model for all receptive fields,and the principal component features and the principal component coefficients distribution were visualized for easy-understanding the PCA results of pesticide on-line-mixing images. The visualization showed a little clustering effect of receptive fields in eight different flowrates. Finally,the uniformity was visualized through the similarity heap maps and the non-uniformity value was calculated according to the feature similarity of the principal components. The experimental results showed that,when the flowrate decreased gradually,the average non-uniformity value increased from 6.492 to 18.262 and the standard deviation increased from 0.536 to 3.625,and the uniformity of pesticide on-line-mixing became worse. The results revealed that the proposed method is a promising method to evaluate uniformity of pesticide on-line-mixing. This method provided powerful technical support for the optimization of the structure of mixers and for the realization of the precision pesticide spraying.
引文
[1]周凤芳.背负式喷雾机农药药水在线混合方法的研究[D].南京:南京林业大学,2012.ZHOU F F. Research of pesticide on-line mixing method for engine powered knapsack sprayer[D]. Nanjing:Nanjing Forestry University,2012.
[2]陈勇,郑加强.精确施药可变量喷雾控制系统的研究[J].农业工程学报,2005,21(5):69-72.CHEN Y,ZHENG J Q. Control system for precision pesticide application based on variable rate technology[J]. Transactions of the CSAE,2005,21(5):69-72.
[3]徐幼林.植保机械混药器及其农药在线混合性能研究[D].南京:南京林业大学,2009.XU Y L. Study on mixers for plant protection machinery and chemical in-line-mixing performance[D]. Nanjing:Nanjing Forestry University,2009.
[4]CAI X,WALGENBACH M,DOERPMOND M,et al. ClosedLoop control of chemical injection rate for a direct nozzle injection system[J]. Sensors,2016,16(1):127.
[5] SHEN Y,ZHU H. Embedded computer controlled premixing inline injection system for air-assisted variable-rate sprayers[J].Transactions of the Asabe,2015,58(1):39-46.
[6]SHEN Y,ZHU H P,LIU H,et al. Development of a real-time chemical injection system for air-assisted variable-rate sprayers[C]∥2013 ASABE Annual International Meeting,Kansas City,Missouri,July 21-24,2013.
[7]VONDRICKA J,HLOBEN P,LAMMERS P S. Optimization of direct nozzle injection system for site-specific herbicide application[C]∥2007 ASABE Annual International Meeting,Minneapolis,Minnesota,June 17-20,2007.
[8]VONDRICKA J,LAMMERS P S. Measurement of mixture homogeneity in direct injection systems[J]. Transactions of the Asabe,2009,52(1):61-66.
[9]ZHU H. Reducing metering lag and nonuniformity for injection sprayers applying highly viscous fluid[D]. Columbus:Ohio State University,1996.
[10] SUDDUTH K A,BORGELT S C,HOU J. Performance of a chemical injection sprayer system[J]. Applied Engineering in Agriculture,1995,11(3):343-348.
[11]徐幼林,郭敬坤,郑加强.农药在线混合均匀度高速摄影分析[J].农业机械学报,2011,42(8):75-79.XU Y L,GUO J K,ZHENG J Q. Mixing uniformity of chemical and water in direct injection system[J]. Transactions of the Chinese Society for Agricultural Machinary,2011,42(8):75-79.
[12]郭敬坤.农药药水在线混合效果的数字图像处理技术研究[D].南京:南京林业大学,2008.GUO J K. Research on mixing degree of direct injection system for pesticide sprayer[D]. Nanjing:Nanjing Forestry University,2008.
[13]宋海潮,徐幼林,郑加强,等.脂溶性农药旋动射流混药器结构分析与混合均匀性试验[J].农业工程学报,2016,32(23):86-92.SONG H C,XU Y L,ZHENG J Q,et al. Structural analysis and mixing uniformity experiments of swirling jet mixer for applying fat-soluble pesticides[J]. Transactions of the CSAE,2016,32(23):86-92.
[14]BERTHIAUX H,MOSOROV V,TOMCZAK L,et al. Principal component analysis for characterising homogeneity in powder mixing using image processing techniques[J]. Chemical Engineering&Processing Process Intensification,2006,45(5):397-403.
[15]CHEN L,WU C,FAN W,et al. Adaptivelocal receptive field convolutional neural networks for handwritten Chinese character recognition[J]. Communications in Computer&Information Science,2014,484:455-463.
[16]BOUREAU Y L,PONCE J,LECUN Y. A theoretical analysis of feature pooling in vision algorithms[C]∥Proc. International Conference on Machine Learning,2010:328-333.
[17]ZHU S,DU J. Visual tracking using max-average pooling and weight-selection strategy[J]. Journal of Applied Mathematics,2014,2014:1-10.
[18]PAGOLA M,FORCEN J I,BARRENECHEA E,et al. A study on the cardinality of ordered average pooling in visual recognition[C]∥Iberian Conference on Pattern Recognition and Image Analysis. Cham:Springer,2017:437-444.
[19]HINTON G E,DAYAN P,REVOW M. Modeling the manifolds of images of handwritten digits[J]. IEEE Transactions on Neural Networks,1997,8(1):65.
[20]GOODFELLOW I,BENGIO Y,COURVILLE A. Deep learning[M].Cambridge:MIT Press,2016:502-504.
[2]陈勇,郑加强.精确施药可变量喷雾控制系统的研究[J].农业工程学报,2005,21(5):69-72.CHEN Y,ZHENG J Q. Control system for precision pesticide application based on variable rate technology[J]. Transactions of the CSAE,2005,21(5):69-72.
[3]徐幼林.植保机械混药器及其农药在线混合性能研究[D].南京:南京林业大学,2009.XU Y L. Study on mixers for plant protection machinery and chemical in-line-mixing performance[D]. Nanjing:Nanjing Forestry University,2009.
[4]CAI X,WALGENBACH M,DOERPMOND M,et al. ClosedLoop control of chemical injection rate for a direct nozzle injection system[J]. Sensors,2016,16(1):127.
[5] SHEN Y,ZHU H. Embedded computer controlled premixing inline injection system for air-assisted variable-rate sprayers[J].Transactions of the Asabe,2015,58(1):39-46.
[6]SHEN Y,ZHU H P,LIU H,et al. Development of a real-time chemical injection system for air-assisted variable-rate sprayers[C]∥2013 ASABE Annual International Meeting,Kansas City,Missouri,July 21-24,2013.
[7]VONDRICKA J,HLOBEN P,LAMMERS P S. Optimization of direct nozzle injection system for site-specific herbicide application[C]∥2007 ASABE Annual International Meeting,Minneapolis,Minnesota,June 17-20,2007.
[8]VONDRICKA J,LAMMERS P S. Measurement of mixture homogeneity in direct injection systems[J]. Transactions of the Asabe,2009,52(1):61-66.
[9]ZHU H. Reducing metering lag and nonuniformity for injection sprayers applying highly viscous fluid[D]. Columbus:Ohio State University,1996.
[10] SUDDUTH K A,BORGELT S C,HOU J. Performance of a chemical injection sprayer system[J]. Applied Engineering in Agriculture,1995,11(3):343-348.
[11]徐幼林,郭敬坤,郑加强.农药在线混合均匀度高速摄影分析[J].农业机械学报,2011,42(8):75-79.XU Y L,GUO J K,ZHENG J Q. Mixing uniformity of chemical and water in direct injection system[J]. Transactions of the Chinese Society for Agricultural Machinary,2011,42(8):75-79.
[12]郭敬坤.农药药水在线混合效果的数字图像处理技术研究[D].南京:南京林业大学,2008.GUO J K. Research on mixing degree of direct injection system for pesticide sprayer[D]. Nanjing:Nanjing Forestry University,2008.
[13]宋海潮,徐幼林,郑加强,等.脂溶性农药旋动射流混药器结构分析与混合均匀性试验[J].农业工程学报,2016,32(23):86-92.SONG H C,XU Y L,ZHENG J Q,et al. Structural analysis and mixing uniformity experiments of swirling jet mixer for applying fat-soluble pesticides[J]. Transactions of the CSAE,2016,32(23):86-92.
[14]BERTHIAUX H,MOSOROV V,TOMCZAK L,et al. Principal component analysis for characterising homogeneity in powder mixing using image processing techniques[J]. Chemical Engineering&Processing Process Intensification,2006,45(5):397-403.
[15]CHEN L,WU C,FAN W,et al. Adaptivelocal receptive field convolutional neural networks for handwritten Chinese character recognition[J]. Communications in Computer&Information Science,2014,484:455-463.
[16]BOUREAU Y L,PONCE J,LECUN Y. A theoretical analysis of feature pooling in vision algorithms[C]∥Proc. International Conference on Machine Learning,2010:328-333.
[17]ZHU S,DU J. Visual tracking using max-average pooling and weight-selection strategy[J]. Journal of Applied Mathematics,2014,2014:1-10.
[18]PAGOLA M,FORCEN J I,BARRENECHEA E,et al. A study on the cardinality of ordered average pooling in visual recognition[C]∥Iberian Conference on Pattern Recognition and Image Analysis. Cham:Springer,2017:437-444.
[19]HINTON G E,DAYAN P,REVOW M. Modeling the manifolds of images of handwritten digits[J]. IEEE Transactions on Neural Networks,1997,8(1):65.
[20]GOODFELLOW I,BENGIO Y,COURVILLE A. Deep learning[M].Cambridge:MIT Press,2016:502-504.