基于图像处理的水稻成熟期密度检测
|
摘要
农业机械化的大力发展可以提高农民的劳动生产率和经济效益。农业机械化水平的提高,不仅有利于丰产丰收,而且有利于减少对环境的污染和农业持续高效地发展。长期以来,农作物生长密度信息的及时准确获取,一直是制约农业机械自动化水平提高的关键因素,也是农业工程研究领域的难题。如稻麦联合收割机工作时,不考虑行走速度的变化,脱粒滚筒的喂入量与稻麦的生长密度成正比。即作物密度越大,喂入量相应增加。但由于土、肥、水、光照等田间作物生长因素分布的随机性,作物生长密度不可能非常均匀,因此喂入量是随机变化的。随机变化的喂入量会影响到脱粒间隙,滚筒线速度,凹板长度,清粮风扇的排风量等一系列可变参数的改变,很难保证联合收割机械工作在一个较理想的工作性能范围。
针对目前联合收割机在收获过程中脱不净,夹带,脱粒损失大等情况,本文从研究水稻成熟期冠层密度分布出发,利用计算机图像处理等技术手段,提出了一种基于RGB彩色模型的水稻冠层密度分布图像的检测方法。
实验地点选择在肥西县丰乐镇水稻产区。实验分两期:早稻—2个品种;晚稻—4个品种。数据采集及处理方法为:(1)样本数据获取:通过限定相同的面积A,采集自然状态下的成熟期水稻图像,然后将水稻从距根部相同高度处割下,脱粒称重,获得水稻毛重,草重等重量参数m。通过计算获得水稻密度值D。不考虑其他次要因素,建立密度值D与喂入量F的函数关系F=f(D)。(2)图像处理过程:综合运用图像处理、植物生理学、色度学、几何特征、距离特征等方面的知识,利用计算机图像处理技术对采集到的彩色图像进行分析比较,寻找R,G,B最优组合,得到2 R +G颜色特征向量图像。对2 R +G颜色特征量图像采用迭代法自动选取阈值将谷、叶从背景中分割出来。2R+G颜色组合值能从彩色图像中最大限度地分割出稻叶等与喂入量直接相关因素,土壤,杂草等背景因素在分割后的图像中全部显示为0。(3)进行数据拟合:综合运用计算机数据处理数据拟合技术,相关性分析技术,显著性检验技术,对冠层图像像素值与水稻喂入量进行数据拟合及相关性检验,最终建立喂入量随像素值变化的检测模型。通过相关性分析及显著性检验可以得出结论:水稻联合收割机喂入量与2R+G颜色特征图的像素值的变化具有显著的正相关性。
本文建立的成熟期水稻喂入量随2R+G颜色特征量变化的检测模型对改进联合收割机的工况有很大的参考价值。为联合收割机在作业过程中实时调整调整喂入量,提高劳动生产率提供一种可行方案。
The strong development of agricultural mechanization can increase productivity of farmers and economic benefits. Raising the level of agricultural mechanization is not only conducive to high-yield harvest, but also will help to reduce environmental pollution and agricultural sustainable and efficient development. A long time, the Information of crop density to obtain timely and accurate information has been constrained the level of agricultural machinery automation of key factors, and also is the field of agricultural engineering research problems. Such as during rice and wheat combine harvester work, not considering the changes in walking speed,threshing cylinder and the volume of the feed density is directly proportional to the growth of rice and wheat。That is, the greater the crop density, the feed volume of a corresponding increases. But because of the random distribution of soil, fertilizer, water, light and other factors of field crops, crop growth can not be very uniform in density, and so feeding volume is random change. Random changes will affect a series of changes such as the feeding threshing gap, roller speed, the length of intaglio and the exhaust volume variable parameters of the fans which purge grain out, it is difficult to ensure that the work of the combine harvester machinery in a ideal performance of the scope of work.
According to the phenomenon of non-net, entrainment, threshing losses, etc in the work course of combine, this paper studies the distribution of rice canopy density, uses computer image processing techniques and raises a detection method of the rice canopy density distribution based on the RGB color images model .
Experimental site selects in the rice producing town of the FengLe town of Feixi. Experiment in two periods: early rice-2varieties; late rice-4varieties. Data acquisition and processing methods are: (1)Sample data acquisition: Through the same limited area of A, we get the rice images under natural state And then cut it from the roots of rice from the same height, weight it , weight it then get the weight of parameters such as gross weight, net weight etc. we get the value of the density of rice by calculating and that does not take into account other factors secondary , The establishment of the density value of the volume of D and F a function of feeding relations F = f (D). (2)Image processing: The integrated use of image processing, plant pathology, Chromaticity, geometric features, from the characteristics of knowledge, using computer image processing technology we can analyze the color images to find the R, G, B the optimal combination, then get the 2R+G color feature vector images. Features of color images using iterative method will be automatically selected threshold Valley to separate leaf from the background. 2R + G color combinations from the color images can segment maximally related directly factors between rice leaves and feeding, such soil, weeds in the background factors all is 0 after the image segmentation.(3)Data fitting: By comprehensive use of computer data processing fitting techniques, the relevance of analytical techniques, a significant test of technology, the image pixel value of the canopy and the amount of rice to feed data fitting and related test, We can carry out data fitting and related test between the pixel value of the canopy Images of rice and the feeding volume of rice and finally establish the detection model that the feeding volume of rice changs with the image pixels.Through correlation analysis and test of significance can be concluded: the volume of feed rice combine harvester and the 2R + G color feature map changes in the value of the pixel has significant positive correlation.
In this paper, the detection model of the feeding volume of the mature rice changing with the 2R + G color images to improve the working condition of combines harvesters has great reference value. This method provides a feasible option which helps combine harvester in operation for the process of feeding real-time adjustment to adjust the volume and improve labor productivity.
针对目前联合收割机在收获过程中脱不净,夹带,脱粒损失大等情况,本文从研究水稻成熟期冠层密度分布出发,利用计算机图像处理等技术手段,提出了一种基于RGB彩色模型的水稻冠层密度分布图像的检测方法。
实验地点选择在肥西县丰乐镇水稻产区。实验分两期:早稻—2个品种;晚稻—4个品种。数据采集及处理方法为:(1)样本数据获取:通过限定相同的面积A,采集自然状态下的成熟期水稻图像,然后将水稻从距根部相同高度处割下,脱粒称重,获得水稻毛重,草重等重量参数m。通过计算获得水稻密度值D。不考虑其他次要因素,建立密度值D与喂入量F的函数关系F=f(D)。(2)图像处理过程:综合运用图像处理、植物生理学、色度学、几何特征、距离特征等方面的知识,利用计算机图像处理技术对采集到的彩色图像进行分析比较,寻找R,G,B最优组合,得到2 R +G颜色特征向量图像。对2 R +G颜色特征量图像采用迭代法自动选取阈值将谷、叶从背景中分割出来。2R+G颜色组合值能从彩色图像中最大限度地分割出稻叶等与喂入量直接相关因素,土壤,杂草等背景因素在分割后的图像中全部显示为0。(3)进行数据拟合:综合运用计算机数据处理数据拟合技术,相关性分析技术,显著性检验技术,对冠层图像像素值与水稻喂入量进行数据拟合及相关性检验,最终建立喂入量随像素值变化的检测模型。通过相关性分析及显著性检验可以得出结论:水稻联合收割机喂入量与2R+G颜色特征图的像素值的变化具有显著的正相关性。
本文建立的成熟期水稻喂入量随2R+G颜色特征量变化的检测模型对改进联合收割机的工况有很大的参考价值。为联合收割机在作业过程中实时调整调整喂入量,提高劳动生产率提供一种可行方案。
The strong development of agricultural mechanization can increase productivity of farmers and economic benefits. Raising the level of agricultural mechanization is not only conducive to high-yield harvest, but also will help to reduce environmental pollution and agricultural sustainable and efficient development. A long time, the Information of crop density to obtain timely and accurate information has been constrained the level of agricultural machinery automation of key factors, and also is the field of agricultural engineering research problems. Such as during rice and wheat combine harvester work, not considering the changes in walking speed,threshing cylinder and the volume of the feed density is directly proportional to the growth of rice and wheat。That is, the greater the crop density, the feed volume of a corresponding increases. But because of the random distribution of soil, fertilizer, water, light and other factors of field crops, crop growth can not be very uniform in density, and so feeding volume is random change. Random changes will affect a series of changes such as the feeding threshing gap, roller speed, the length of intaglio and the exhaust volume variable parameters of the fans which purge grain out, it is difficult to ensure that the work of the combine harvester machinery in a ideal performance of the scope of work.
According to the phenomenon of non-net, entrainment, threshing losses, etc in the work course of combine, this paper studies the distribution of rice canopy density, uses computer image processing techniques and raises a detection method of the rice canopy density distribution based on the RGB color images model .
Experimental site selects in the rice producing town of the FengLe town of Feixi. Experiment in two periods: early rice-2varieties; late rice-4varieties. Data acquisition and processing methods are: (1)Sample data acquisition: Through the same limited area of A, we get the rice images under natural state And then cut it from the roots of rice from the same height, weight it , weight it then get the weight of parameters such as gross weight, net weight etc. we get the value of the density of rice by calculating and that does not take into account other factors secondary , The establishment of the density value of the volume of D and F a function of feeding relations F = f (D). (2)Image processing: The integrated use of image processing, plant pathology, Chromaticity, geometric features, from the characteristics of knowledge, using computer image processing technology we can analyze the color images to find the R, G, B the optimal combination, then get the 2R+G color feature vector images. Features of color images using iterative method will be automatically selected threshold Valley to separate leaf from the background. 2R + G color combinations from the color images can segment maximally related directly factors between rice leaves and feeding, such soil, weeds in the background factors all is 0 after the image segmentation.(3)Data fitting: By comprehensive use of computer data processing fitting techniques, the relevance of analytical techniques, a significant test of technology, the image pixel value of the canopy and the amount of rice to feed data fitting and related test, We can carry out data fitting and related test between the pixel value of the canopy Images of rice and the feeding volume of rice and finally establish the detection model that the feeding volume of rice changs with the image pixels.Through correlation analysis and test of significance can be concluded: the volume of feed rice combine harvester and the 2R + G color feature map changes in the value of the pixel has significant positive correlation.
In this paper, the detection model of the feeding volume of the mature rice changing with the 2R + G color images to improve the working condition of combines harvesters has great reference value. This method provides a feasible option which helps combine harvester in operation for the process of feeding real-time adjustment to adjust the volume and improve labor productivity.
引文
[1]农业机械产品何时能让农民放心?http://www.people.com.cn
[2]中国消费者协会2002年第7号消费警示http://www.hlqingan.gov.cn
[3]邵陆寿,魏雅媚.基于GA与BP结合的收获机械脱粒性能建模与仿真[J].系统仿真学报, 2003 ,15(9):1294～1296.
[4]崔怀雷.基于模糊控制的变量施药系统的研究[D].安徽农业大学硕士论文,2004.
[5] Han YJ ,Hayes J C. Soil cover determination using color image analysis[J] . Transaction of the ASAE ,1990 ,33(4) :1402～1407.
[6] Tarbell K A,Reid J F.A computer vision system for characterizing corn growth and development[J].Transaction of the ASAE,1991,34(5) :2245～2255.
[7] Gonzalez R C ,Woods R E. Digital image processing[M]. Addison Wesley Publishing Co. Inc. ,1992.
[8] Shearer S A ,Thomasson J A ,Mcneill S G. Filter selection for NIR sensing of plant and soil materials[J] . Transaction of the ASAE ,1996 ,39(3) :1209～1214.
[9] Hayes J C ,Han YJ . Comparison of crop cover measyring system[A] . ASAE paper[C] ,St . Joseph ,Mich. ,1989.
[10] Meyer G E ,Mehta T , Kocher M F ,et al. Textural imaging and discriminant analysis for distinguishing weeds for spot spraying[J].Transaction of the ASAE ,1998 ,41(4) :1189～1197.
[11] Tian L ,Slaughter D C ,Norris R F. Outdoor field machine vision identification of tomato seedlings for automated weed control[J].Transaction of the ASAE ,1997 ,40(6) :1761～1768.
[12] Tang L ,Tian L ,Steward B L. Color image segmentation with genetic algorithm for in2field weed sensing[J] . Transaction of the ASAE ,2000 ,43(4) :1019～1027.
[13] EI-Faki M S ,Zhang N ,Peterson D E. Factors affecting color based weed detection[J] . Transaction of the ASAE ,2000 ,43 (4) :1001～1009.
[14] Guyer D E ,Miles G E ,Schreiber M M ,et al. Machine vision and image processing for plant identification[J] . Trans of the ASAE ,1986 ,29(6) :1500～1507.
[15] Weobbecke D M ,Meyer G E ,Von Bargen K,et al. Shape features for identifying young weeds using image analysis[J].Transaction of the ASAE ,1995 ,38(1) :271～281.
[16] Zhang N ,Chaisattapagon C. Effective criteria for weed identification in wheat fields using machine vision[J]. Transaction of the ASAE ,1995 ,38(3) :965～974.
[17] Cardina J ,Sparrow D H ,McCoy E L. Analysis of spatial distribution of common lambsquarters in soybean [J] .Weed Sci. ,1995 ,43(2) :258～268.
[18]赵先贵.植物花粉图像识别系统研究[J].西北林学院学报,1999,14(4):31～34.
[19]孙德敏.基于纹理的植物叶片红外图像的模式识别[J].模式识别与人工智能,2004,17(1):78～81.
[20]金开正.杂草密度与作物产量损失预测的一种新方法[J].安徽农业科学, 1999,27(5):454～456.
[21]邵陆寿.水稻生长密度图像处理与仿真[J].系统仿真学报.论文全文附后
[22]应义彬.机械视觉技术在农产品收获和加工自动化中的应用[J].农业机械学报,2000,31(3)112～115.
[23] Rafael c.Gonzalez,Richard E.Woods[M].Digital Image Processing.电子工业出版社,2003.
[24]张志涌.精通MATLAB6.5版[M].北京:北京航空航天大学出版社,2003.
[25]尹泽明,丁春利.精通MATLAB6[M].北京:清华大学出版社,2002.
[26] (美)Rafael C.Gonzalez,Richard E.Woods,Steven L.Eddins.数字图像处理(MATLAB版)[M].北京:电子工业出版社,2004.
[27]徐飞,施晓红.MATLAB应用图像处理[M].西安:西安电子科技大学出版社,2002.
[28]罗军辉,冯平,哈力旦·A等.MATLAB7.0在图像处理中的应用[M].北京:机械工业出版社,2005.
[29]吕朝辉等.机器视觉田间植物检测与识别技术[J].吉林工业大学自然科学学报,2001,31(3):90～94.
[30]章毓晋.图像处理与分析[M].北京:清华大学出版社,1999.
[31] (美)Rafael C.Gonzalez,Richard E.Woods阮秋琦,阮宇智等译.数字图像处理[M].北京:电子工业出版,2003.
[32]吕朝辉.用BP神经网络进行秧苗图像分割[J].农业工程学报,2001,17(3): 146~148.
[33]葛婧.基于计算机图像处理技术的作物病害等级检测[硕士学位论文]合肥:安徽农业大学2007
[34]戴之祥水稻及玉米植株密度的可见光图像检测[硕士学位论文]合肥:安徽农业大学2006
[35] [40]Rafael C Gonzalez,Richard E Woods,阮秋琦等译.数字图像处理[M].北京:电子工业出版社,2003,485～486
[36]刘宗杰,徐凤君,毕守东等.概率论与数理统计.上海:上海交通大学,1998, 237~261.
[37]袁志发,周静芋.多元统计分析[M].北京:科学出版社,2002,75~187.
[38]洪楠,林爱华,李志辉等.SPSS for Windows统计分析教程.北京:电子工业出版社,2000.
[39]马庆国等,应用统计学:数理统计方法、数据获取与SPSS应用科学出版社2005年11月版
[40]苏金明,傅荣华,周建斌等.统计软件SPSS for Windows实用指南.北京:电子工业出版社,2000,439~443.
[41]唐启义、冯明光等实用统计分析及其DPS数据处理系统北京:科学出版社,2002年5月版
[42]唐启义、冯明光等实用统计分析及其计算机处理平台北京:中国农业出版社,1997年9月版
[43]周仁郁等统计方法与实验成都:西南交通大学出版社2002.4
[44]方萍,何延等实验设计与统计杭州:浙江大学出版社,2003.6
[45]朱海涛,陈黎等DPS数据处理系统在药学研究数据处理中的应用[J].医药导报, 2006,(04) .
[46]王毅,汪洋,蔡华.基于小波变换的逐步线性回归分析预测大坝位移[J].绘信息与工程, 2008,(04) .
[47]王国锋,王子良,王太勇,王双利. Matlab在时间序列分析中的应用[J].应用科技, 2003,(05) .
[2]中国消费者协会2002年第7号消费警示http://www.hlqingan.gov.cn
[3]邵陆寿,魏雅媚.基于GA与BP结合的收获机械脱粒性能建模与仿真[J].系统仿真学报, 2003 ,15(9):1294～1296.
[4]崔怀雷.基于模糊控制的变量施药系统的研究[D].安徽农业大学硕士论文,2004.
[5] Han YJ ,Hayes J C. Soil cover determination using color image analysis[J] . Transaction of the ASAE ,1990 ,33(4) :1402～1407.
[6] Tarbell K A,Reid J F.A computer vision system for characterizing corn growth and development[J].Transaction of the ASAE,1991,34(5) :2245～2255.
[7] Gonzalez R C ,Woods R E. Digital image processing[M]. Addison Wesley Publishing Co. Inc. ,1992.
[8] Shearer S A ,Thomasson J A ,Mcneill S G. Filter selection for NIR sensing of plant and soil materials[J] . Transaction of the ASAE ,1996 ,39(3) :1209～1214.
[9] Hayes J C ,Han YJ . Comparison of crop cover measyring system[A] . ASAE paper[C] ,St . Joseph ,Mich. ,1989.
[10] Meyer G E ,Mehta T , Kocher M F ,et al. Textural imaging and discriminant analysis for distinguishing weeds for spot spraying[J].Transaction of the ASAE ,1998 ,41(4) :1189～1197.
[11] Tian L ,Slaughter D C ,Norris R F. Outdoor field machine vision identification of tomato seedlings for automated weed control[J].Transaction of the ASAE ,1997 ,40(6) :1761～1768.
[12] Tang L ,Tian L ,Steward B L. Color image segmentation with genetic algorithm for in2field weed sensing[J] . Transaction of the ASAE ,2000 ,43(4) :1019～1027.
[13] EI-Faki M S ,Zhang N ,Peterson D E. Factors affecting color based weed detection[J] . Transaction of the ASAE ,2000 ,43 (4) :1001～1009.
[14] Guyer D E ,Miles G E ,Schreiber M M ,et al. Machine vision and image processing for plant identification[J] . Trans of the ASAE ,1986 ,29(6) :1500～1507.
[15] Weobbecke D M ,Meyer G E ,Von Bargen K,et al. Shape features for identifying young weeds using image analysis[J].Transaction of the ASAE ,1995 ,38(1) :271～281.
[16] Zhang N ,Chaisattapagon C. Effective criteria for weed identification in wheat fields using machine vision[J]. Transaction of the ASAE ,1995 ,38(3) :965～974.
[17] Cardina J ,Sparrow D H ,McCoy E L. Analysis of spatial distribution of common lambsquarters in soybean [J] .Weed Sci. ,1995 ,43(2) :258～268.
[18]赵先贵.植物花粉图像识别系统研究[J].西北林学院学报,1999,14(4):31～34.
[19]孙德敏.基于纹理的植物叶片红外图像的模式识别[J].模式识别与人工智能,2004,17(1):78～81.
[20]金开正.杂草密度与作物产量损失预测的一种新方法[J].安徽农业科学, 1999,27(5):454～456.
[21]邵陆寿.水稻生长密度图像处理与仿真[J].系统仿真学报.论文全文附后
[22]应义彬.机械视觉技术在农产品收获和加工自动化中的应用[J].农业机械学报,2000,31(3)112～115.
[23] Rafael c.Gonzalez,Richard E.Woods[M].Digital Image Processing.电子工业出版社,2003.
[24]张志涌.精通MATLAB6.5版[M].北京:北京航空航天大学出版社,2003.
[25]尹泽明,丁春利.精通MATLAB6[M].北京:清华大学出版社,2002.
[26] (美)Rafael C.Gonzalez,Richard E.Woods,Steven L.Eddins.数字图像处理(MATLAB版)[M].北京:电子工业出版社,2004.
[27]徐飞,施晓红.MATLAB应用图像处理[M].西安:西安电子科技大学出版社,2002.
[28]罗军辉,冯平,哈力旦·A等.MATLAB7.0在图像处理中的应用[M].北京:机械工业出版社,2005.
[29]吕朝辉等.机器视觉田间植物检测与识别技术[J].吉林工业大学自然科学学报,2001,31(3):90～94.
[30]章毓晋.图像处理与分析[M].北京:清华大学出版社,1999.
[31] (美)Rafael C.Gonzalez,Richard E.Woods阮秋琦,阮宇智等译.数字图像处理[M].北京:电子工业出版,2003.
[32]吕朝辉.用BP神经网络进行秧苗图像分割[J].农业工程学报,2001,17(3): 146~148.
[33]葛婧.基于计算机图像处理技术的作物病害等级检测[硕士学位论文]合肥:安徽农业大学2007
[34]戴之祥水稻及玉米植株密度的可见光图像检测[硕士学位论文]合肥:安徽农业大学2006
[35] [40]Rafael C Gonzalez,Richard E Woods,阮秋琦等译.数字图像处理[M].北京:电子工业出版社,2003,485～486
[36]刘宗杰,徐凤君,毕守东等.概率论与数理统计.上海:上海交通大学,1998, 237~261.
[37]袁志发,周静芋.多元统计分析[M].北京:科学出版社,2002,75~187.
[38]洪楠,林爱华,李志辉等.SPSS for Windows统计分析教程.北京:电子工业出版社,2000.
[39]马庆国等,应用统计学:数理统计方法、数据获取与SPSS应用科学出版社2005年11月版
[40]苏金明,傅荣华,周建斌等.统计软件SPSS for Windows实用指南.北京:电子工业出版社,2000,439~443.
[41]唐启义、冯明光等实用统计分析及其DPS数据处理系统北京:科学出版社,2002年5月版
[42]唐启义、冯明光等实用统计分析及其计算机处理平台北京:中国农业出版社,1997年9月版
[43]周仁郁等统计方法与实验成都:西南交通大学出版社2002.4
[44]方萍,何延等实验设计与统计杭州:浙江大学出版社,2003.6
[45]朱海涛,陈黎等DPS数据处理系统在药学研究数据处理中的应用[J].医药导报, 2006,(04) .
[46]王毅,汪洋,蔡华.基于小波变换的逐步线性回归分析预测大坝位移[J].绘信息与工程, 2008,(04) .
[47]王国锋,王子良,王太勇,王双利. Matlab在时间序列分析中的应用[J].应用科技, 2003,(05) .