光谱变换方法对黑土养分含量高光谱遥感反演精度的影响
2018-10-18张东辉赵英俊赵宁博杨越超
张东辉,赵英俊,秦 凯,赵宁博,杨越超
光谱变换方法对黑土养分含量高光谱遥感反演精度的影响
张东辉,赵英俊,秦 凯,赵宁博,杨越超
(核工业北京地质研究院 遥感信息与图像分析技术国家级重点实验室,北京 100029)
高光谱遥感反演黑土养分含量时,光谱变换方法对提取精度具有显著影响,为明确二者响应关系,提高反演精度和稳定度,该文以黑龙江建三江地区为研究区,引入航空高光谱成像系统CASI-1500,获取380~1 050 nm数据进行分析。均匀采样60个样品,化验获得其有机质、全氮、全磷和全钾含量数据,利用神经网络方法对有机质含量、支持向量机对氮、磷、钾含量进行建模。对比研究了重采样(RE)、对数倒数(LR)、一阶微分(FD)、包络线去除(CR)和多元散射校正(MSC)变换5种光谱变换后的提取精度。结果表明:MSC、MSC、LR和RE光谱变换方法分别应用到有机质、氮、磷和钾特征波段的组合运算中,得出黑土养分含量的空间分布精度相对最高,预测样本的决定系数分别为0.748、0.673、0.631和0.420。
遥感;土壤;模型;光谱变换法;神经网络;支持向量机
0 引 言
随着高光谱遥感技术在生态评价领域的研究深入,开展光谱遥感反演与地球化学验证,建立黑土养分快速评价技术体系,能为黑土资源管理提供科学依据。根据土壤不同养分的跃迁能级差不同,研究物质吸收光谱曲线,得出物质的各组成成分[1]。现实研究中,由于土壤的理化性质、上覆状况和环境扰动千差万别,导致光谱特征和成分含量的对应关系难以准确建立。需要在大量可靠光谱数据积累的基础上,通过统计学习方法,逐步发现这些对应关系,并在与实测结果综合分析的基础上,解释对应关系的作用原理。
机载高光谱遥感在获取光谱数据同时,采集了高精度的空间数据,使得研究土壤多种成分的空间分布关系成为可能,进而能够计算出物质间赋存和转运关系。由于直接从土壤光谱中提取稀有元素的困难性,这种赋存关系的掌握,将为高光谱在这一领域研究的拓展提供技术手段。在获取土壤光谱后,需要经过光谱异常筛选、平滑去噪、重采样、光谱变换和光谱定量化计算等处理方法,而其中的光谱变换方法,能够起到增强有价值波段信息,提高建模精度的作用[2]。光谱变换的目的是通过将原始反射率进行转换,形成一系列反射率自变量,这种自变量能够放大或者缩小特征峰的反射率值,提升光谱识别的概率[3]。在与理化成分分析数据建立回归模型时,经过多种方法的综合验证,分析光谱数据和化验数据的匹配关系[4]。
何挺等对土壤光谱进行了14种变换,研究了土壤光谱反射特性与有机质之间的关系,证明反射率对数的一阶微分对土壤有机质含量最为敏感[5]。刘焕军等[6]通过对典型黑土可见光/近红外波段光谱反射特性研究,得出归一化变换可以部分消除不同土样测试过程中存在的噪声。Andreas Steinberg等通过对不同有机质含量土壤的光谱曲线吸收特征进行分析,得出包络线去除和反射率的倒数的对数处理建立的偏最小二乘回归模型预测效果最佳[7]。方少文等研究表明土壤全氮与一阶微分转换后反射率相关系数较高的峰值位置在820、1 400、1 430、1 630、1 800、1 930 nm等波段[8]。
黑土光谱在可见/近红波段范围内反射率普遍较低,吸收特征不显著,且易受水分和秸秆等因素的干扰,直接使用测量光谱所建立的反演模型的推广性受到限制[9]。本文以黑龙江建三江地区黑土样本为研究对象,对黑土光谱进行重采样、对数倒数、一阶微分、包络线去除和多元散射校正等变换,建立了其有机质、氮、磷、钾等养分含量的定量提取模型,通过对比模型预测值与实测值的误差,对5种光谱变换方法的适用性进行了研究,以期为光谱变换方法的选择提供科学参考。
1 材料与方法
1.1 研究区概况
研究区位于黑龙江省建三江地区,系黑龙江、松花江、乌苏里江汇流的河间地带。以盛产绿色优质水稻闻名,故有“中国绿色米都”之誉。地势低平,地形标高50~60 m。由黄土状粉质黏土、淤泥质粉质粘土构成,主要分布于山前台地顶部[10]。腐殖质富集,加之母质黏重,水不能迅速下渗,缓慢淋滤形成黑土层[11]。表层为黑色腐殖质层,厚30~60 cm,最厚可达1m以上,多具圆柱状或粒状结构;其下为质地黏重的淀积层,棕色铁锰结核一般较多,再下为棕黄色粘性母质层[12]。
1.2 数据来源
数据由CASI-1500航空高光谱成像光谱系统(加拿大ITRES)获取。光谱范围为380~1 050 nm,空间分辨率为1.5 m,连续光谱通道数55,光谱带宽10 nm,总视场角40°,瞬时视场角0.028°,每行像元数1 470,绝对辐射精度<2%。飞行高度3 km(图1)。地面测量铺设黑白布,采用ASD Field Spec光谱仪获取定标光谱,光谱范围为350~2 500 nm,采集光谱分辨率为1 nm。
1.3 土壤样采集与化学测定
研究区长9.27 km,宽5.36 km,面积约50 km2。采样点60个,样本1的坐标为132.747°E,47.232°N,样本60的坐标为132.857°E,47.272°N,按0.75km间隔采集土样,采样时间为飞行作业同步采样。测区表层为黑色腐殖质层,厚30~60 cm,最厚可达1 m以上,多具圆柱状或粒状结构。当天同步采集表层0~20 cm的土样,剔除大的植物残茬、石砺等杂物,置于实验室风干研磨,过0.15 mm筛选用于土壤养分含量测定。有机质采用重铬酸钾容量-外加热法测定,全氮、全磷和全钾含量分别采用凯氏定氮法、NaOH碱熔钼锑抗比色法和钾火焰原子吸收分光光度法测定含量[5]。土壤养分含量测定结果中,样本1~45用于训练集,其余15个样本用于预测(表1)。
图1 研究区及样点布置
表1 不同样本点土壤养分含量信息表
1.4 光谱变换方法
选用R语言klap包实现支持向量机模型[13],AMORE包实现BP神经网络的建立,重采样采用Mathlab实现,航空高光谱波段运算由ENVI 5.3的bandmath实现。选用5种光谱变换算法试验[14]。
1)重采样(resampling,RE)
针对黑土光谱与养分含量提取的尺度不确定性问题[15],通过重采样能够确定最佳的提取波长间隔。计算公式为
式中D为采样间隔;=D(D为偶数);=D+1(D为奇数)。
2)对数倒数(logarithmic reciprocal,LR)
光谱通过对数计算后,能够成为相对值较近似的值,避免数据过大或过小[16]。倒数将这一新值转换为同一量级的数据,使之更具备可对比性[17]。计算公式为
式中Rnew_i为光谱变换后的新值;R为原始光谱反射率(下同)。
3)一阶微分(first derivative,FD)
通过对反射光谱进行数据模拟,计算不同阶的微分值迅速确定光谱变化点及最大最小反射率的波长位置。一阶微分增强了光谱变化和压缩的影响[18]。计算公式为
式中R+Di为与原始波段间隔一定范围的光谱反射率;D为波长的间隔,视变换需要而定。
4)包络线去除(continuum removal,CR)
包络线去除可以有效突出光谱曲线的吸收和反射特征,并将反射率归一化到0~1[19]。计算过程为:对光谱曲线上的所有“凸”出峰值点,比较大小,得到最大值点,作为包络线的一个端点,计算该点与长波方向各个极大值点连线的斜率,以斜率最大点作为下一个包络线端点进行循环,直至最后一点;再以最大值点作为包络线端点,向短波方向进行类似计算,以斜率最小点为下一端点进行循环,直到曲线开始点;沿波长增加方向连接这些端点,即形成包络线。
5)多元散射校正(multivariate scattering correction,MSC)
多元散射校正可以有效地消除散射影响,增强与成分含量相关的光谱吸收信息[20]。首先取所有光谱的平均光谱作为标准光谱,将每个样品光谱与标准光谱进行一元线性回归运算,计算各光谱相对于标准光谱的回归常数和系数,减去线性平移量,同时除以回归系数修正光谱的基线相对倾斜,达到对每个光谱的基线平移和偏移都在标准光谱的参考下予以修正的目的,在不损失光谱吸收信息的前提下,提高了光谱的信噪比。计算公式为
2 结果与分析
2.1 养分含量与光谱关系分析
2.1.1 不同含量的黑土养分光谱特征
将60个黑土样本按养分含量大小排序,分析在可见-近红波段范围内光谱变换规律[21]。一是通过光谱特性与含量的机理分析,有机质和氮元素的光谱特征较为明显,而磷和钾含量与光谱反射率整体的走势关系不显著;二是所选取的60个采样点,有机质和氮元素含量建模样本区分度较好,标准偏差达到0.23和0.09,而全磷和全钾的标准偏差仅为0.03和0.02,微小的含量差异导致较难得出回归系数较好的模型。试验结果也表明,标准偏差越好,所建模型的回归系数就越高。鉴于论文重点研究光谱变换方法对四种养分提取的影响,而建立精度更高回归系数数学模型不是论文的研究重点,在相同数学模型下,横向对比四种光谱变换方法是有一定意义的[22]。
图2为不同含量的黑土养分光谱特征图。每个区间范围取2条光谱曲线进行分析,得出随着有机质含量增高,黑土反射率逐渐降低(图2a)。其中,8号样品有机质达到4.46 g/kg,反射率显著低于其他样品;而41号和53号样品有机质质量分数在3.3 g/kg左右,其反射率明显高于总体光谱均值。当有机质含量较低时,由于土壤含水量和混合像元等干扰,这一规律会逐渐减弱,直至不显著。氮变化规律是与有机质光谱曲线类似,随着氮含量增高,反射率逐渐降低(图2b)。其中,9号和50号样品氮质量分数高于2.28 g/kg,反射率低于其他样品。而随着氮元素含量的进一步减少,这一规律不显著。由于磷元素含量相对较小,在光谱曲线上的反射特征不明显,在可见-近红光谱范围内的变换没有显著的规律(图2c)。同样,钾元素含量在可见-近红光谱范围内的变换也没有显著的规律(图2d)。
图2 不同黑土养分含量的光谱特征
2.1.2 土壤主要养分特征波段提取
对60个采样点不同养分含量进行逐波段求反射率对养分的相关系数[23-25](图3)。
结果表明,与其他土壤养分相比,有机质各个波段相关系数最高,均值达到0.39,氮和磷相关系数接近,分别为0.28和0.30,钾相关系数最低,为0.05。选取相关系数较高的前5个波段,作为建模波段[26]。有机质为933.6、914.5、905、866.8和943.1 nm,氮为933.6、866.8、876.3、847.7和914.5 nm,磷为950、933.6、866.8、857.3和914.5 nm,钾为523.7、771.5、571.4、695.3和533.2 nm。
图3 逐波段光谱反射率与黑土养分含量的相关关系
2.2 变换方法对养分含量预测的影响
2.2.1 黑土养分含量预测方法
对黑土光谱分别进行重采样(RE)、对数倒数(LR)、一阶微分(FD)、包络线去除(CR)和多元散射校正(MSC)变换等共计5种光谱数据[27]。对比了神经网络、支持向量机和偏最小二乘法对4种养分的提取精度,有机质和全钾信息提取精度最高的算法是神经网络法,误差分别为1.21%和0.81%,而支持向量机算法在提取全氮和全磷信息时,验证样本的实测均值和预测均值完全吻合,精度最高。因此,选用神经网络法,对研究区内所有航空高光谱数据进行有机质和全钾信息提取。采用支持向量机方法,对研究区内全氮和全磷信息进行建模和提取[28]。
具体参数设置为:支持向量机模型类别选eps-regression,核函数选linear线性,采用试错法计算最佳gamma和惩罚因子,gamma设置为10-5~10-1,惩罚因子选10、50和100,根据20遍交叉检验方式评价每次组合的错误偏差[29]。所建神经网络模型为一个3层神经网络,即5-3-1,含1个隐层,完成预测模型的建立。神经元学习率为4,采用最小均方根误差法设置训练误差函数,隐藏层神经元激励函数为传递函数tansig,输出层神经元激励函数为线性函数purelin,训练权值更新方法为含有动量的自适应梯度下降法ADAPTgdwm[30]。
2.2.2 重采样评估光谱尺度效应
理论上光谱分辨率越高,土壤养分特征波段越显著,模型反演的精度越高[31]。而实际提取中,将多个波段进行合成,能够降低噪声的干扰,提高模型的鲁棒性[11]。因此,需要评估每种土壤养分提取的最佳光谱分辨率。将高光谱数据采样为55、44、33、22和11个波段,提取特征波段的反射率,进行黑土养分提取。以15个预测样本均方根误差RMSE和模型决定系数2作为尺度效应评估指标,RMSE越小,说明模型的预测精度越高,2越大,模型的稳定性越好[32]。
通过对比不同重采样光谱的反演结果,在5种重采样方法中,波段数55所建立的模型,均方根RMSE相对都最小或持平,而且模型决定系数2均是最高或持平,说明波段数的增多,能够在一定程度上提升模型反演的精度。
2.2.3 建立响应关系模型
将原始光谱反射率集处理为重采样RE、对数倒数LR、一阶微分FD、包络线去除CR和多元散射校正MSC反射率新值(图4),利用神经网络方法对60个样本的有机质含量进行建模,利用支持向量机对60个样本的氮、磷、钾含量进行建模,得出其模型预测精度[33](表2)。
建模样本中,有机质、氮、磷和钾光谱变换精度最高的方法分别是MSC(0.922)、MSC(0.872)、LR(0.621)和RE(0.423);预测样本中,有机质、氮、磷和钾光谱变换精度排序与建模样本一致,分别为MSC(0.748)、MSC(0.673)、LR(0.631)和RE(0.420)。建模样本和预测样本的均方根RMSE呈现出一致的排序规律,表明有机质和全氮选择MSC变换方法,而全磷和全钾在LR和RE变换下,具有最高的模型决定系数和最低的均方根误差。
图4 黑土光谱的RE、LR、FD、CR和MSC处理结果(1号样本点)
表2 不同光谱变换方法的土壤养分建模结果
2.3 提取结果
依次将决定系数较高的MSC、MSC、LR和RE光谱变换方法应用到有机质、氮、磷和钾特征波段的组合运算中,得出黑土养分含量的空间分布情况(图5)。分析得出,研究区黑土养分含量空间分布呈现明显的地块规律,这与这一地区农业开发较为成熟有关。不同的地块由不同的农户种植,对地块施肥、秸秆处理和灌溉休耕的处理各不相同,导致黑土养分的差异。总体上研究区有机质和全氮分布规律近似,呈现出相似的分布规律。而磷元素和钾元素由于含量较低,提取的误差较大。
图5 采用最佳光谱变换后的黑土养分含量(g·kg-1)提取空间分布图
3 结 论
为提高光谱反演精度,将原始光谱反射率数据处理为重采样RE、对数倒数LR、一阶微分FD、包络线去除CR和多元散射校正MSC等变换值。利用神经网络方法对60个样本的有机质含量进行建模,利用支持向量机对60个样本的氮、磷、钾含量进行建模。MSC、MSC、LR和RE光谱变换方法分别应用到有机质、氮、磷和钾特征波段的组合运算中,预测样本的决定系数分别为0.748、0.673、0.631和0.420,得出黑土养分含量的空间分布精度相对最高。得出了每种黑土养分提取精度最佳的变换方法,以及五种光谱变换方法的提取精度差异,对于掌握光谱变换与黑土养分含量响应关系提供了定量依据。
[1] 史舟. 土壤地面高光谱遥感原理与方法[M]. 北京:科学出版社,2014.
[2] Daniel Žížala, Tereza Zádorová, Jiří Kapička. Assessment of soil degradation by erosion based on analysis of soil properties using aerial hyperspectral images and ancillary data[J]. Remote Sens, 2017, 9(1): 28-40.
[3] 周鼎浩,薛利红,李颖,等. 基于可见–近红外光谱的水稻土全磷反演研究[J]. 土壤,2014,46(1):47-52.
Zhou Dinghao, Xue Lihong, Li Ying, et al. Visible–near infrared reflectance spectroscopy for prediction of total phosphorus content in paddy soil[J]. Soil, 2014, 46(1): 47-52. (in Chinese with English abstract)
[4] 刘焕军,潘越,窦欣,等. 黑土区田块尺度土壤有机质含量遥感反演模型[J]. 农业工程学报,2018,34(1):127-133.
Liu Huanjun, Pan Yue, Dou Xin, et al. Soil organic matter content inversion model with remote sensing image in field scale of black soil area[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2018, 34(1): 127-133. (in Chinese with English abstract)
[5] 何挺,王静,林宗坚,等. 土壤有机质光谱特征研究[J]. 武汉大学学报:信息科学版,2006,31(11):975-979.
He Ting, Wang Jing, Lin Zongjian, et al. Spectral features of soil organic matter[J]. Geomatics and Information Science of Wuhan University, 2006, 31(11): 975-979. (in Chinese with English abstract)
[6] 刘焕军,张柏,赵军,等. 黑土有机质含量高光谱模型研究[J]. 土壤学报,2007,44(1):27-32.
Liu Huanjun, Zhang Bai, Zhao Jun, et al. Spectral models for prediction of organic matter in black soil[J]. Acta Pedologica Sinica, 2007, 44(1): 27-32. (in Chinese with English abstract)
[7] Andreas Steinberg, Sabine Chabrillat, Antoine Stevens, et al. Prediction of common surface soil properties based on vis–nir airborne and simulated enmap imaging spectroscopy data: prediction accuracy and influence of spatial resolution[J]. Remote Sens, 2016, 8(7): 613-627.
[8] 方少文,杨梅花,赵小敏,等. 红壤区土壤有机质光谱特征与定量估算:以江西省吉安县为例[J]. 土壤学报,2014,51(5):1003-1010. Fang Shaowen, Yang Meihua, Zhao Xiaomin, et al. Spectral characteristics and quantitative estimation of som in red soil typical of Ji’an county, Jiangxi Province[J]. Acta Pedologica Sinica, 2014, 51(5): 1003-1010. (in Chinese with English abstract)
[9] 薛利红,周鼎号,李颖,等. 不同利用方式下土壤有机质和全磷的可见近红外高光谱反演[J]. 土壤学报,2014,51(5):993-1001.
Xue Lihong, Zhou Dinghao, Li Ying, et al. Prediction of soil organic matter and total phosphorus with vis-nir hyperspectral inversion relative to land use[J]. Acta Pedologica Sinica, 2014, 51(5): 993-1001. (in Chinese with English abstract)
[10] 张俊华,马天成,贾科利. 典型龟裂碱土土壤光谱特征影响因素研究[J]. 农业工程学报,2014,30(23):158-165.
Zhang Junhua, Ma Tiancheng, Jia Keli. Factors affecting spectral characteristics of typical takyr solonetzs[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2014, 30(23): 158-165. (in Chinese with English abstract)
[11] Jin Xiuliang, Song Kaishan, Du Jia, et al. Comparison of different satellite bands and vegetation indices for estimation of soil organic matter based on simulated spectral configuration [J]. Agricultural and Forest Meteorology, 2017, 244: 57-71.
[12] 马驰. 基于Landsat 8吉林中北部地区土壤有机质定量反演研究[J]. 干旱区资源与环境,2017,31(2):167-172.
Ma Chi. Study on the quantitative retrieval of soil organic matter in the north and middle part of Jilin based on Landsat 8[J]. Journal of Arid Land Resources and Environment, 2017, 31(2): 167-172. (in Chinese with English abstract)
[13] 李萍,赵庚星,高明秀,等. 黄河三角洲土壤含水量状况的高光谱估测与遥感反演[J]. 土壤学报,2015,52(6):1262-1271.
Li Ping, Zhao Gengxing, Gao Mingxiu, et al. Hyperspectral estimation and remote sensing retrieval of soil water regime in the Yellow River delta[J]. Acta Pedologica Sinica, 2015, 52(6): 1262-1271. (in Chinese with English abstract)
[14] 吕杰,郝宁燕,史晓亮. 基于流形学习的土壤高光谱数据特征提取研究[J]. 干旱区资源与环境,2015,29(7):176-180.
Lü Jie, Hao Ningyan, Shi Xiaoliang. Extraction of hyperspectral characteristics of soil based on manifold learning[J]. Journal of Arid Land Resources and Environment, 2015, 29(7): 176-180. (in Chinese with English abstract)
[15] Zhang P, Li Y. Study on the comparisons of the establishment of two mathematical modeling methods for soil organic matter content based on spectral reflectance[J]. Spectroscopy and Spectral Analysis, 2016, 36(3): 903-910.
[16] 王昶,黄驰超,徐光辉,等. 近红外光谱结合偏最小二乘法快速评估土壤质量[J]. 土壤学报,2013,50(5):36-45.
Wang Chuang, Huang Chichao, Xu Guanghui, et al. Rapid evaluation of soil quality through a near infrared–partial least squares (NIR-PLS) method[J]. Acta Pedologica Sinica, 2013, 50(5): 36-45. (in Chinese with English abstract)
[17] Zhao H, Feng X, Xiao P. Contour extraction of green cover along urban roads from remote sensing imagery based on frequency domain features[J]. Remote Sensing Information, 2014, 29(3): 50-56.
[18] Doustfatemeh I, Baleghi Y. Comprehensive urban area extraction from multispectral medium spatial resolution remote-sensing imagery based on a novel structural feature[J]. International Journal of Remote Sensing, 2016, 37(18): 4225-4242.
[19] Andreas Steinberg, Sabine Chabrillat, Antoine Stevens, et al. Prediction of common surface soil properties based on Vis-NIR airborne and simulated EnMAP imaging spectroscopy data: Prediction Accuracy and Influence of Spatial Resolution[J]. Remote Sens, 2016, 8(7): 613-627.
[20] 李硕,汪善勤,张美琴,等. 基于可见长丘红外光谱比较主成分回归、偏最小二乘回归和反向传播神经网络对土壤氮的预测研究[J]. 光学学报,2012,32(8):0830001-0830005.
Li Shuo, Wang Shanqin, Zhang Meiqin, et al. Comparison among principal component regression,partial least squares regression and back propagation neural network for prediction of soil nitrogen with visible–near infrared spectroscopy[J]. Acta Optica Sinica, 2012, 32(8): 0830001-0830005. (in Chinese with English abstract)
[21] 徐永明,蔺启忠,王璐,等. 基于高分辨率反射光谱的土壤营养元素估算模型[J]. 土壤学报,2006,43(5):709-716.
Xu Yongming, Lin Qizhong, Wang Lu, et al. Model for estimating soil nutrient elements based on high resolution reflectance spectra[J]. Acta Pedologica Sinica, 2006, 43(5): 709-716. (in Chinese with English abstract)
[22] Yu L, Liu X B, Liu G Z, et al. Experiment research and analysis of spectral prediction on soil leaking oil content (In Chinese)[J]. Spectroscopy and Spectral Analysis, 2016, 36(4): 1116-1120.
[23] 洪长桥,郑光辉,陈昌春. 苏北滨海土壤碳酸钙含量反射光谱估算研究[J]. 土壤学报,2016,53(5):1120-1128.
Hong Changqiao, Zheng Guanghui, Chen Changchun. Estimation of CaCO3content in coastal soil of north Jiangsu with reflectance spectroscopy[J]. Acta Pedologica Sinica, 2016, 53(5): 1120-1128. (in Chinese with English abstract)
[24] 王凯龙,熊黑钢,张芳. 基于高光谱数据预测土壤碱化程度最佳模型及其影响因素的研究[J]. 土壤,2014,46(3):544-549.
Wang Kailong, Xiong Heigang, Zhang Fang. Optimal model of soil pH and influencing factors by using hyperspectral data[J]. Soils, 2014, 46(3): 544-549. (in Chinese with English abstract)
[25] 高灯州,曾从盛,章文龙,等. 闽江口湿地土壤全氮含量的高光谱遥感估算[J]. 生态学杂志,2016,35(4):952-959.
Gao Dengzhou, Zeng Congsheng, Zhang Wenlong, et al. Estimating of soil total nitrogen concentration based on hyperspectral remote sensing data in Minjiang River estuarine wetland[J]. Chinese Journal of Ecology, 2016, 35(4): 952-959. (in Chinese with English abstract)
[26] 汪善勤,舒宁,蔡崇法,等. 基于PCA和SRRI的潮土土壤属性与田间光谱关系研究[J]. 土壤,2008,40(6):960-970.
Wang Shanqin, Shu Ning, Cai Chongfa, et al. Relationship between in situ spectroscopy and properties of fluvo-aquic soil based on PCA and SRRI[J]. Soils, 2008, 40(6): 960-970. (in Chinese with English abstract)
[27] 金慧凝,张新乐,刘焕军,等. 基于光谱吸收特征的土壤含水量预测模型研究[J]. 土壤学报,2016,53(3):627-634.
Jin Huining, Zhang Xinle, Liu Huanjun, et al. Soil moisture predicting model based on spectral absorption characteristics of the soil[J]. Acta Pedologica Sinica, 2016, 53(3): 627-634. (in Chinese with English abstract)
[28] 张东辉,赵英俊,秦凯. 一种新的光谱参量预测黑土养分含量模型[J]. 光谱学与光谱分析,2018,38(9):1-5. Zhang Donghui, Zhao Yingjun, Qin Kai. A new model for predicting black soil nutrient content by spectral parameters[J]. Spectroscopy and Spectral Analysis, 2018, 38(9): 1-5. (in Chinese with English abstract)
[29] 李晨,张国伟,周治国,等. 滨海盐土土壤水分的高光谱参数及估测模型[J]. 应用生态学报,2016,27(2):525-531.
Li Chen,Zhang Guowei,Zhou Zhiguo,et al. Hyperspectral parameters and prediction model of soil moisture in coastal saline[J]. Chinese Journal of Applied Ecology 2016, 27(2): 525-531. (in Chinese with English abstract)
[29] 柴思跃,马维玲,刘高焕,等. GA-PLS方法提取土壤水盐光谱特征的精度分析[J]. 遥感技术与应用,2015,30(4):638-644.
Chai Siyue, Ma Weiling, Liu Gaohuan, et al. Accuracy analysis of GA-PLS based soil water salinity hyperspectral characteristics mining approach[J]. Remote Sensing Technology and Application, 2015, 30(4): 638-644. (in Chinese with English abstract)
[31] 程先锋,宋婷婷,陈玉,等. 滇西兰坪铅锌矿区土壤重金属含量的高光谱反演分析[J]. 岩石矿物学杂志,2017,36(1):60-69.
Cheng Xianfeng, Song Tingting, Chen Yu, et al. Retrieval and analysis of heavy metal content in soil based on measured spectra in the Lanping Zn-Pb mining area, western Yunnan Province[J]. Acta Petrologica ET Mineralogica, 2017, 36(1): 60-69. (in Chinese with English abstract)
[32] 任红艳. 宝山矿区农田土壤水稻系统重金属污染的遥感监测[B]. 南京:南京农业大学,2008.
Ren Hongyan. Use of hyperimultiple spectral data in monitoring heavy metal pollution in soil-paddy plant system nearby Baoshan mines[J]. Nanjing Agricultural University, 2008. (in Chinese with English abstract)
[33] 陈松超,冯来磊,李硕,等. 基于局部加权回归的土壤全氮含量可见-近红外光谱反演[J]. 土壤学报,2015,52(2):312-319.
Chen Songchao, Feng Lailei, Li Shuo, et al. Vis-nir spectral inversion for prediction of soil total nitrogen content in laboratory based on locally weighted regression[J]. Acta Pedologica Sinica, 2015, 52(2): 312-319. (in Chinese with English abstract)
Influence of spectral transformation methods on nutrient content inversion accuracy by hyperspectral remote sensing in black soil
Zhang Donghui, Zhao Yingjun, Qin Kai, Zhao Ningbo, Yang Yuechao
(100029,)
In order to improve the precision and stability of the soil nutrient content inversion model in black soil area, taking Jiansanjiang area in Heilongjiang province as the study area, and the airborne hyperspectral imaging system CASI-1500 (380-1 050 nm) as the analysis data, the influence of different spectral transformation methods on the accuracy was researched. 60 samples were evenly sampled, and the contents of organic matter, total nitrogen, total phosphorus and total potassium were obtained through laboratory tests. The content of organic matter was determined by potassium dichromate capacity external heating method. The content of total nitrogen, total phosphorus and total potassium was determined by Kjeldahl method, NaOH alkali antimony colorimetric method and potassium flame atomic absorption spectrophotometry. 60 black soil samples were sorted according to nutrient content, and the spectral transformation in the visible near red range was analyzed. The change rule of organic matter is that the reflectance decreases with the increase of content. The change rule of nitrogen is similar to the spectral curve of organic matter. With the increase of nitrogen content, the reflectance decreases. The transformation of phosphorus and potassium in the visible near red spectrum is not significant. The nutrient correlation coefficients of 60 samples at different sampling points were calculated by spectral reflectance. The results show that the correlation coefficient of each band is the highest, the mean value is 0.39, the correlation coefficients of nitrogen and phosphorus are close to 0.28 and 0.30, and the correlation coefficient of potassium is the lowest, which is 0.05. The first 5 bands with high correlation coefficient are selected as modeling bands, that of organic matter is 933.6, 914.5, 905, 866.8 and 943.1 nm, and that of nitrogen is 933.6, 866.8, 876.3, 847.7 and 914.5 nm. The content of organic matter and support vector machine were used to model nitrogen, phosphorus and potassium contents. The extraction accuracies of 5 spectral transformations which are resampling (RE), logarithmic reciprocal (LR), first order derivative (FD), continuum removal (CR) and multivariate scatter correction (MSC) transformation are compared. The most accurate methods for the spectral transformation of organic matter, nitrogen, phosphorus and potassium are MSC, MSC, LR and RE, respectively. Five spectral transformation methods are used to calculate the2of each model, and the order of modeling accuracy for soil organic matter prediction is MSC (0.922) > RE (0.529) > LR (0.432) > CR (0.414) > FD (0.018). The modeling accuracy of multiple scattering correction transformation is significantly higher than that of the other four methods. The order of prediction accuracy or total phosphorus is MSC (0.872) > CR (0.387) > RE (0.256) > LR (0.029) > FD (0.012), and the prediction accuracy of the multivariate scattering correction transformation is also the highest. The highest prediction accuracies of total phosphorus and total potassium are LR (0.621) and RE (0.423). In turn, the MSC, MSC, LR and RE spectral transformation methods with high coefficient of determination are applied to the combined operation of the characteristics of organic matter, nitrogen, phosphorus and potassium, and the spatial distribution of nutrient content in black soil is obtained. The results show that the spectral transformation methods of MSC, MSC, LR and RE are applied to calculate soil organic matter, nitrogen, phosphorus and potassium, respectively, the spatial distribution accuracy of nutrient content in black soil is the highest, and the determination coefficients of predicted samples are 0.748, 0.673, 0.631 and 0.420, respectively.
remote sensing; soils; models; spectral transformation methods; neural networks; support vector machines
10.11975/j.issn.1002-6819.2018.20.018
TP79
A
1002-6819(2018)-20-0141-07
2018-03-07
2018-09-03
国家自然科学基金项目(41602333)、“十三五”装备预先研究专项技术项目(32101080302)、遥感信息与图像分析技术国家级重点实验室重点基金(9140C720105140C72001)和中国地质调查局项目(12120113073000)联合资助
张东辉,博士,高级工程师,主要从事高光谱遥感技术与应用研究。Email:donghui222@163.com
张东辉,赵英俊,秦 凯,赵宁博,杨越超. 光谱变换方法对黑土养分含量高光谱遥感反演精度的影响[J]. 农业工程学报,2018,34(20):141-147. doi:10.11975/j.issn.1002-6819.2018.20.018 http://www.tcsae.org
Zhang Donghui, Zhao Yingjun, Qin Kai, Zhao Ningbo, Yang Yuechao. Influence of spectral transformation methods on nutrient content inversion accuracy by hyperspectral remote sensing in black soil[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2018, 34(20): 141-147. (in Chinese with English abstract) doi:10.11975/j.issn.1002-6819.2018.20.018 http://www.tcsae.org