基于可见光—近红外图像的幼龄檀香全磷含量诊断
|
摘要
【目的】檀香是一种典型珍贵树种,在幼龄期时,不合理的田间施肥会影响其正常生长,降低存活率。因此,本文提出了一种基于可见光-近红外图像的幼龄檀香全磷营养诊断方法,为实时监测珍贵树种生长状态及养分需求提供参考。【方法】通过将野外获取的檀香图像转换到HSI颜色空间,提取S和I通道,利用二者在使用Otsu分割后产生的优势互补,并结合形态学运算,从复杂背景中提取出檀香。计算出形状、纹理和光谱及植被指数特征后,分别使用显著性分析(ST)和平均影响值(MIV)方法进行变量筛选,并使用遗传算法(GA)初始化BP神经网络的权值和阈值,最终得到预测结果。【结果】(1)复杂背景下的檀香分割中,S通道和I通道相结合可以将大部分背景(天空、土壤、其他绿色植物)与目标檀香分割开,同时结合7×7中值滤波、形态学运算和超G因子,将其他毛刺去除。与常用的支持向量机相比,本文提出的分割算法结果更接近于目视解译,像素数和颜色误差更小。(2)对不同施磷水平下各特征进行分析发现,适当增加施磷量有利于促进叶绿素的形成,使得纹理更均匀清晰,加快叶片生长;当过量时则会破坏叶绿体,造成叶片组织出现变化,导致叶片黄化,叶片出现网状脉纹,增加了纹理复杂程度。(3) ST与MIV筛选出的变量差异较大,通过GA-BPNN训练结果可知,MIV方法筛选出的变量对全磷含量的影响更大,预测集得到的决定系数达到0. 801,平均残差为0. 032 g/kg,均方根误差为0. 666 g/kg。【结论】通过处理可见光-近红外图像,实现了幼龄檀香的全磷含量诊断,有效提高了磷肥利用率,同时也可以减小过量施肥引起的地下水污染等生态问题。
[Objective]Sandalwood is a typical precious tree species. During its young stage,more or less fertilization will affect its growth and reduce survival rate. In this paper,a total phosphorus nutrition diagnosis method for young sandalwood based on visible light and near infrared image recognition is proposed. It provides a reference for real-time monitoring of the growth state and nutrient requirements of precious tree species. [Method]S and I channels were extracted after converting field acquired sandalwood images to HSI color space. By combing the advantage of S and I channels segmentationresults using Otsu method and morphological operation,sandalwood was extracted from the complex background. We used different methods to optimize BP neural network. On one hand,ST and MIV methods were used to select the variables in shape,texture,spectrum and vegetation index. On the other hand,genetic algorithm( GA) was used to initialize the weights and thresholds,and the prediction results were finally obtained. [Results]( 1) In the complex background of sandalwood segmentation,the combination of H channel and S channel successfully separated most of the background( sky,soil and other green plants) from target sandalwood. Median filter sized in 7 × 7,morphological operation and super G factor were used to remove other burrs.( 2) The characteristics under different levels of phosphorus application showed that the appropriate increase fertilizer could promote the compound of chlorophyll,make the texture more uniform and clear,and increase the growth of the leaves. When the application exceeded the best value,the chloroplast could be destroyed,texture changed and leaf color turned into yellow.( 3) The variables selected by ST and MIV were different. GA-BPNN training results showed that the variables selected by the MIV method had greater influence on the total phosphorus content. The determinant coefficient of the prediction set was 0. 801,the mean residual was 0. 032 g/kg,and the root mean square error was 0. 666 g/kg. [Conclusion] In this paper,the total phosphorus content of young sandalwood was predicted by processing the visible light and near infrared image. By this method,the utilization rate of phosphate fertilizer will be improved effectively,and the ecological problems such as groundwater pollution caused by excessive fertilization could also be reduced.
[Objective]Sandalwood is a typical precious tree species. During its young stage,more or less fertilization will affect its growth and reduce survival rate. In this paper,a total phosphorus nutrition diagnosis method for young sandalwood based on visible light and near infrared image recognition is proposed. It provides a reference for real-time monitoring of the growth state and nutrient requirements of precious tree species. [Method]S and I channels were extracted after converting field acquired sandalwood images to HSI color space. By combing the advantage of S and I channels segmentationresults using Otsu method and morphological operation,sandalwood was extracted from the complex background. We used different methods to optimize BP neural network. On one hand,ST and MIV methods were used to select the variables in shape,texture,spectrum and vegetation index. On the other hand,genetic algorithm( GA) was used to initialize the weights and thresholds,and the prediction results were finally obtained. [Results]( 1) In the complex background of sandalwood segmentation,the combination of H channel and S channel successfully separated most of the background( sky,soil and other green plants) from target sandalwood. Median filter sized in 7 × 7,morphological operation and super G factor were used to remove other burrs.( 2) The characteristics under different levels of phosphorus application showed that the appropriate increase fertilizer could promote the compound of chlorophyll,make the texture more uniform and clear,and increase the growth of the leaves. When the application exceeded the best value,the chloroplast could be destroyed,texture changed and leaf color turned into yellow.( 3) The variables selected by ST and MIV were different. GA-BPNN training results showed that the variables selected by the MIV method had greater influence on the total phosphorus content. The determinant coefficient of the prediction set was 0. 801,the mean residual was 0. 032 g/kg,and the root mean square error was 0. 666 g/kg. [Conclusion] In this paper,the total phosphorus content of young sandalwood was predicted by processing the visible light and near infrared image. By this method,the utilization rate of phosphate fertilizer will be improved effectively,and the ecological problems such as groundwater pollution caused by excessive fertilization could also be reduced.
引文
[1]刘小金,徐大平,杨曾奖,等.海南尖峰岭檀香心材比例、精油含量和成分的分布特征[J].华南农业大学学报,2016,37(5):66-71.Liu X J,Xu D P,Yang Z J,et al.Heartwood proportion and distribution of essential oil content and composition of Santalum album in Jianfeng Mountain Hainan[J].Journal of South China Agricultural University,2016,37(5):66--71.
[2]林芬芳,丁晓东,付志鹏,等.基于互信息理论的水稻磷素营养高光谱诊断[J].光谱学与光谱分析,2009,29(19):2467-2470.Lin F F,Ding X D,Fu Z P,et al.Application of mutual information to variable selection in diagnosis of phosphorus nutrition in rice[J].Spectroscopy and Spectral Analysis,2009,29(19):2467-2470.
[3]王艳丽,王京,刘国顺,等.磷施用量对烤烟根系生理及叶片光合特性的影响[J].植物营养与肥料学报,2016,22(2):410-417.Wang Y L,Wang J,Liu G S,et al.Effects of different phosphorus levels on root physiological and leaf photosynthetic characteristics of flue-cured tobacco[J].Journal of Plant Nutrition and Fertilizer,2016,22(2):410--417.
[4]连慧达,裴红宾,张永清,等.施磷量对不同品种红小豆形态和生理特性的影响[J].植物营养与肥料学报,2015,21(3):792-799.Lian H D,Pei H B,Zhang Y Q,et al.Effect of phosphorus fertilization on morphological and physiological characteristics of adzuki beans[J].Journal of Plant Nutrition and Fertilizer,2015,21(3):792-799.
[5]郑亚萍,信彩云,王才斌,等.磷肥对花生根系形态、生理特性及产量的影响[J].植物生态学报,2013,37(8):777-785.Zheng Y P,Xin C Y,Wang C B,et al.Effects of phosphorus fertilizer on root morphology,physiological characteristics and yield in peanut(Arachis hypogaea)[J].Chinese Journal of Plant Ecology,2013,37(8):777--785.
[6]赵霞,徐大平,杨曾奖,等.养分胁迫对降香黄檀幼苗生长及叶片养分状况的影响[J].生态学杂志,2017,36(6):1503-1508.Zhao X,Xu D P,Yang Z J,et al.Effects of nutrient stress on seedling growth and foliar nutrient status of Dalbergia odorifera[J].Chinese Journal of Ecology,2017,36(6):1503--1508.
[7]李美清,李晋阳,毛罕平.基于光谱特征和生理特征的番茄磷营养诊断方法[J].农业机械学报,2016,47(3):286-291.Li M Q,Li J Y,Mao H P.Tomatoes phosphorus nutrition diagnosis based on spectral and physiological characteristics[J].Transactions of The Chinese Society of Agricultural Machinery,2016,47(3):286--291.
[8]何彩莲,郑顺林,万年鑫,等.马铃薯光谱及数字图像特征参数对氮素水平的响应及其应用[J].光谱学与光谱分析,2016,36(9):2930--2936.He C L,Zheng S L,Wan N X,et al.Potato spectrum and the digital image feature parameters on the response of the nitrogen level and its application[J].Spectroscopy and Spectral Analysis,2016,36(9):2930-2936.
[9]陈敏,郑曙峰,刘小玲,等.基于数码图像识别的棉花氮营养诊断研究[J].农学学报,2017,7(7):77-83.Chen M,Zheng S F,Liu X L,et al.Cotton nitrogen nutrition diagnosis based on digital image[J].Journal of Agriculture,2017,7(7):77--83.
[10]Osborne S L,Schepers J S,Francis D D,et al.Detection of phosphorus and nitrogen deficiencies in corn using spectral radiance measurements[J].Agronomy Journal,2002,94(6):1215-1221.
[11]黄双萍,岳学军,洪添胜.基于高光谱的柑橘叶片磷含量估算模型实验[J].农业工程学报,2013,44(4):202--207.Huang S P,Yue X J,Hong T S.Hyperspectral estimation model of total phosphorus content for citrus leaves[J].The Chinese Society of Agricultural Machinery,2013,44(4):202-207.
[12]高洪燕,毛罕平,张晓东.生菜叶中磷含量的光谱定量分析[J].农业机械学报,2014,45(增刊):276--280.Gao H Y,Mao H P,Zhang X D.Quantitative determination of phosphorus in lettuce leaf using spectroscopy[J].The Chinese Society of Agricultural Machinery,2014,45(Suppl.):276-280.
[13]Tucker C J.Red and photographic infrared linear combinations for monitoring vegetation[J].Remote Sensing of Environment,1979,8:127-150.
[14]Buschmann C,Nagel E.In vivo spectroscopy and internal optics of leaves as basis for remote sensing of vegetation[J].International Journal of Remote Sensing,1993,14:711--722.
[15]Gitelson A A,Kaufman Y J,Merzlyak M N.Use of green channel in remote sensing of global vegetation from EOS-MODIS[J].Remote Sensing of Environment,1996,58:289--298.
[16]Roujean J L,Breon F M.Estimating PAR absorbed by vegetation from bidirectional reflectance measurements[J].Remote Sensing of Environment,1995,51:375--384.
[17]Gitelson A A,Gritz Y,Merzlyak M N.Relationships between leaf chlorophyll content and spectral reflectance and algorithms for nondestructive chlorophyll assessment in higher plant leaves[J].Journal of Plant Physiology,2003,160:271--282.
[18]Gitelson A A.Wide dynamic range vegetation index for remote quantification of biophysical characteristics of vegetation[J].Journal of Plant Physiology,2004,161:165--173.
[19]Chen J M.Evaluation of vegetation indices and a modified simple ratio for boreal applications[J].Canadian Journal of Remote Sensing,1996,22:229--242.
[20]Datt B.Visible/near infrared reflectance and chlorophyll content in eucalyptus leaves[J].International Journal of Remote Sensing,1999,20:2741--2759.
[21]Wang W,Yao X,Yao X F,et al.Estimating leaf nitrogen concentration with three-band vegetation indices in rice and wheat[J].Field Crops Res,2012,129:90--98.
[22]Mutanga O,Adam E,Adjorlolo C,et al.Evaluating the robustness of models developed from field spectral data in predicting African grass foliar nitrogen concentration using WorldView-2 image as an independent test dataset[J].International Journal of Applied Earth Observation and Geoinformation,2015,34:178--187.
[23]张恒德,张庭玉,李涛,等.基于BP神经网络的污染物浓度多模式集成预报[J].中国环境科学,2018,38(4):1243-1256.Zhang H D,Zhang T Y,Li T,et al.Forecast of air quality pollutants’concentrations based on BP neural network multi-model ensemble method[J].China Environmental Science,2018,38(4):1243--1256.
[24]李哲,张飞,陈丽华,等.光谱指数的植物叶片叶绿素含量估算模型[J].光谱学与光谱分析,2018,38(5):1533--1539.Li Z,Zhang F,Chen L H,et al.Research on spectrum variance of vegetation leaves and estimation model for leaf chlorophyll content based on the spectral index[J].Spectroscopy and Spectral Analysis,2018,38(5):1533--1539.
[25]刘丽颖,官冬杰,杨清伟,等.基于MIV-BP型的喀斯特地区水资源安全影响因素分析[J].水土保持通报,2017,37(5):128-134.Liu L Y,Guan D J,Yang Q W,et al.Influence factors of water resource security in karst area based on MIV-BP model[J].Bulletin of Soil and Water Conservation,2017,37(5):128-134.
[26]张婉婉,杨可明,汪国平,等.基于EMD-SD光谱的玉米叶片叶绿素含量GA-BP模型反演[J].浙江农业学报,2016,28(8):1297--1303.Zhang W W,Yang K M,Wang G P,et al.Study on GA-BPinversing modeling method of corn leaf chlorophyll content based on EMD and spectral derivative method[J].Acta Agriculture Zhejiangnsis,2016,28(8):1297-1303.
[27]杨梅花,赵小敏.基于可见-近红外光谱变量选择的土壤全氮含量估测研究[J].中国农业科学,2014,47(12):2374-2383.Yang M H,Zhao X M.Study on soil total N estimation by VisNIR spectra with variable selection[J].Scientia Agriculture Sinica,2014,47(12):2374--2383.
[28]迟德霞,张伟,王洋.基于EXG因子的水稻秧苗图像分割[J].安徽农业科学,2012(36):17902--17903.Chi D X,Zhang W,Wang Y.Segmentation of rice seedling image based on EXG factor[J].Journal of Anhui Agricultural Science,2012(36):17902-17903.
[29]苏义鑫,夏慧雯.用于风电功率预测的RPCL优化神经网络模型[J].北京工业大学学报,2016,42(5):674-678.Su Y X,Xia H W.Rival penalized competitve learning-based neural network model for wind power forecasting[J].Journal of Beijing University of Technology,2016,42(5):674-678.
[2]林芬芳,丁晓东,付志鹏,等.基于互信息理论的水稻磷素营养高光谱诊断[J].光谱学与光谱分析,2009,29(19):2467-2470.Lin F F,Ding X D,Fu Z P,et al.Application of mutual information to variable selection in diagnosis of phosphorus nutrition in rice[J].Spectroscopy and Spectral Analysis,2009,29(19):2467-2470.
[3]王艳丽,王京,刘国顺,等.磷施用量对烤烟根系生理及叶片光合特性的影响[J].植物营养与肥料学报,2016,22(2):410-417.Wang Y L,Wang J,Liu G S,et al.Effects of different phosphorus levels on root physiological and leaf photosynthetic characteristics of flue-cured tobacco[J].Journal of Plant Nutrition and Fertilizer,2016,22(2):410--417.
[4]连慧达,裴红宾,张永清,等.施磷量对不同品种红小豆形态和生理特性的影响[J].植物营养与肥料学报,2015,21(3):792-799.Lian H D,Pei H B,Zhang Y Q,et al.Effect of phosphorus fertilization on morphological and physiological characteristics of adzuki beans[J].Journal of Plant Nutrition and Fertilizer,2015,21(3):792-799.
[5]郑亚萍,信彩云,王才斌,等.磷肥对花生根系形态、生理特性及产量的影响[J].植物生态学报,2013,37(8):777-785.Zheng Y P,Xin C Y,Wang C B,et al.Effects of phosphorus fertilizer on root morphology,physiological characteristics and yield in peanut(Arachis hypogaea)[J].Chinese Journal of Plant Ecology,2013,37(8):777--785.
[6]赵霞,徐大平,杨曾奖,等.养分胁迫对降香黄檀幼苗生长及叶片养分状况的影响[J].生态学杂志,2017,36(6):1503-1508.Zhao X,Xu D P,Yang Z J,et al.Effects of nutrient stress on seedling growth and foliar nutrient status of Dalbergia odorifera[J].Chinese Journal of Ecology,2017,36(6):1503--1508.
[7]李美清,李晋阳,毛罕平.基于光谱特征和生理特征的番茄磷营养诊断方法[J].农业机械学报,2016,47(3):286-291.Li M Q,Li J Y,Mao H P.Tomatoes phosphorus nutrition diagnosis based on spectral and physiological characteristics[J].Transactions of The Chinese Society of Agricultural Machinery,2016,47(3):286--291.
[8]何彩莲,郑顺林,万年鑫,等.马铃薯光谱及数字图像特征参数对氮素水平的响应及其应用[J].光谱学与光谱分析,2016,36(9):2930--2936.He C L,Zheng S L,Wan N X,et al.Potato spectrum and the digital image feature parameters on the response of the nitrogen level and its application[J].Spectroscopy and Spectral Analysis,2016,36(9):2930-2936.
[9]陈敏,郑曙峰,刘小玲,等.基于数码图像识别的棉花氮营养诊断研究[J].农学学报,2017,7(7):77-83.Chen M,Zheng S F,Liu X L,et al.Cotton nitrogen nutrition diagnosis based on digital image[J].Journal of Agriculture,2017,7(7):77--83.
[10]Osborne S L,Schepers J S,Francis D D,et al.Detection of phosphorus and nitrogen deficiencies in corn using spectral radiance measurements[J].Agronomy Journal,2002,94(6):1215-1221.
[11]黄双萍,岳学军,洪添胜.基于高光谱的柑橘叶片磷含量估算模型实验[J].农业工程学报,2013,44(4):202--207.Huang S P,Yue X J,Hong T S.Hyperspectral estimation model of total phosphorus content for citrus leaves[J].The Chinese Society of Agricultural Machinery,2013,44(4):202-207.
[12]高洪燕,毛罕平,张晓东.生菜叶中磷含量的光谱定量分析[J].农业机械学报,2014,45(增刊):276--280.Gao H Y,Mao H P,Zhang X D.Quantitative determination of phosphorus in lettuce leaf using spectroscopy[J].The Chinese Society of Agricultural Machinery,2014,45(Suppl.):276-280.
[13]Tucker C J.Red and photographic infrared linear combinations for monitoring vegetation[J].Remote Sensing of Environment,1979,8:127-150.
[14]Buschmann C,Nagel E.In vivo spectroscopy and internal optics of leaves as basis for remote sensing of vegetation[J].International Journal of Remote Sensing,1993,14:711--722.
[15]Gitelson A A,Kaufman Y J,Merzlyak M N.Use of green channel in remote sensing of global vegetation from EOS-MODIS[J].Remote Sensing of Environment,1996,58:289--298.
[16]Roujean J L,Breon F M.Estimating PAR absorbed by vegetation from bidirectional reflectance measurements[J].Remote Sensing of Environment,1995,51:375--384.
[17]Gitelson A A,Gritz Y,Merzlyak M N.Relationships between leaf chlorophyll content and spectral reflectance and algorithms for nondestructive chlorophyll assessment in higher plant leaves[J].Journal of Plant Physiology,2003,160:271--282.
[18]Gitelson A A.Wide dynamic range vegetation index for remote quantification of biophysical characteristics of vegetation[J].Journal of Plant Physiology,2004,161:165--173.
[19]Chen J M.Evaluation of vegetation indices and a modified simple ratio for boreal applications[J].Canadian Journal of Remote Sensing,1996,22:229--242.
[20]Datt B.Visible/near infrared reflectance and chlorophyll content in eucalyptus leaves[J].International Journal of Remote Sensing,1999,20:2741--2759.
[21]Wang W,Yao X,Yao X F,et al.Estimating leaf nitrogen concentration with three-band vegetation indices in rice and wheat[J].Field Crops Res,2012,129:90--98.
[22]Mutanga O,Adam E,Adjorlolo C,et al.Evaluating the robustness of models developed from field spectral data in predicting African grass foliar nitrogen concentration using WorldView-2 image as an independent test dataset[J].International Journal of Applied Earth Observation and Geoinformation,2015,34:178--187.
[23]张恒德,张庭玉,李涛,等.基于BP神经网络的污染物浓度多模式集成预报[J].中国环境科学,2018,38(4):1243-1256.Zhang H D,Zhang T Y,Li T,et al.Forecast of air quality pollutants’concentrations based on BP neural network multi-model ensemble method[J].China Environmental Science,2018,38(4):1243--1256.
[24]李哲,张飞,陈丽华,等.光谱指数的植物叶片叶绿素含量估算模型[J].光谱学与光谱分析,2018,38(5):1533--1539.Li Z,Zhang F,Chen L H,et al.Research on spectrum variance of vegetation leaves and estimation model for leaf chlorophyll content based on the spectral index[J].Spectroscopy and Spectral Analysis,2018,38(5):1533--1539.
[25]刘丽颖,官冬杰,杨清伟,等.基于MIV-BP型的喀斯特地区水资源安全影响因素分析[J].水土保持通报,2017,37(5):128-134.Liu L Y,Guan D J,Yang Q W,et al.Influence factors of water resource security in karst area based on MIV-BP model[J].Bulletin of Soil and Water Conservation,2017,37(5):128-134.
[26]张婉婉,杨可明,汪国平,等.基于EMD-SD光谱的玉米叶片叶绿素含量GA-BP模型反演[J].浙江农业学报,2016,28(8):1297--1303.Zhang W W,Yang K M,Wang G P,et al.Study on GA-BPinversing modeling method of corn leaf chlorophyll content based on EMD and spectral derivative method[J].Acta Agriculture Zhejiangnsis,2016,28(8):1297-1303.
[27]杨梅花,赵小敏.基于可见-近红外光谱变量选择的土壤全氮含量估测研究[J].中国农业科学,2014,47(12):2374-2383.Yang M H,Zhao X M.Study on soil total N estimation by VisNIR spectra with variable selection[J].Scientia Agriculture Sinica,2014,47(12):2374--2383.
[28]迟德霞,张伟,王洋.基于EXG因子的水稻秧苗图像分割[J].安徽农业科学,2012(36):17902--17903.Chi D X,Zhang W,Wang Y.Segmentation of rice seedling image based on EXG factor[J].Journal of Anhui Agricultural Science,2012(36):17902-17903.
[29]苏义鑫,夏慧雯.用于风电功率预测的RPCL优化神经网络模型[J].北京工业大学学报,2016,42(5):674-678.Su Y X,Xia H W.Rival penalized competitve learning-based neural network model for wind power forecasting[J].Journal of Beijing University of Technology,2016,42(5):674-678.