高光谱小波能量特征估测土壤有机质含量

引用本文

章涛, 于雷, 易军, 聂艳, 周勇. 高光谱小波能量特征估测土壤有机质含量[J]. 光谱学与光谱分析, 2019,39(10): 3217-3222.
ZHANG Tao, YU Lei, YI Jun, NIE Yan, ZHOU Yong. Determination of Soil Organic Matter Content Based on Hyperspectral Wavelet Energy Features[J]. Spectroscopy and Spectral Analysis, 2019,39(10): 3217-3222.
Doi:10.3964/j.issn.1000-0593(2019)10-3217-06 复制到剪切板

Permissions

光谱学与光谱分析所有

高光谱小波能量特征估测土壤有机质含量

章涛^1,^2,³, 于雷^1,^2,^3,^*, 易军^1,^2,³, 聂艳^1,^2,³, 周勇^1,^2,³

1. 华中师范大学地理过程分析与模拟湖北省重点实验室, 湖北武汉 430079

2. 华中师范大学城市与环境科学学院, 湖北武汉 430079

3. 华中师范大学可持续发展研究中心, 湖北武汉 430079

*通讯联系人 e-mail: yulei@mail.ccnu.edu.cn

作者简介: 章涛, 1994年生, 华中师范大学城市与环境科学学院硕士研究生 e-mail: zhangt@mails.ccnu.edu.cn

收稿日期: 2018-08-27 接受日期: 2018-12-21

基金: 国家自然科学基金项目(41401232), 中央高校基本科研业务费项目(CCNU19QN050), 和湖北省自然科学基金创新群体项目(2016CFA027)资助

摘要

土壤高光谱在采集过程中难以避免噪声干扰, 造成高光谱数据信噪比较低, 影响土壤有机质含量估测精度。尝试探究小波能量特征方法, 降低高光谱噪声, 提升土壤有机质含量高光谱估测模型性能。选取湖北省潜江市运粮湖管理区为试验区, 于2016年9月采集80份深度为0~20 cm的水稻土样本; 土壤样本经风干、碾磨、过筛等一系列处理后, 在实验室内采集样本光谱, 并通过重铬酸钾-外加热法测定土壤有机质含量; 利用浓度梯度法, 将总体样本集(80个样本)划分为建模集(54个样本)和验证集(26个样本); 以mexh为小波基函数进行连续小波变换(continuous wavelet transformation), 将土壤高光谱转换为10个分解尺度的小波系数(wavelet coefficients); 逐尺度计算小波系数的均方根作为小波能量特征(energy features), 将10个尺度的小波能量特征组成小波能量特征向量(energy features vector); 逐尺度逐波长计算小波系数与有机质含量的相关系数, 将达到极显著水平( p<0.01)的小波系数作为敏感小波系数(sensitive wavelet coefficients); 利用主成分分析法(principal component analysis)分别计算土壤高光谱和小波能量特征向量的各主成分载荷, 通过比较两者第一主成分贡献率的高低和两者前三个主成分得分的空间离散程度, 判断小波能量特征转换前后建模自变量的主成分信息变化趋势; 基于小波能量特征向量和敏感小波系数分别建立多元线性回归和偏最小二乘回归土壤有机质含量估测模型。结果表明, 土壤有机质含量越高, 全波段反射率越低, 但不同土样的光谱反射率曲线特征相似, 近红外部分的反射率(780~2 400 nm)高于可见光部分(400~780 nm); 敏感小波系数对应的波长为494, 508, 672, 752, 1 838和2 302 nm; 土壤高光谱与小波能量特征向量的第一主成分贡献率分别为96.28%和99.13%, 小波能量特征向量的前三个主成分散点较土壤高光谱的主成分散点在空间上更为聚集, 表明小波能量特征方法有效减少了噪声影响; 比较全部土壤有机质含量估测模型, 以小波能量特征向量为自变量的多元线性回归模型具有最佳估测精度, 其验证集决定系数( R²)、相对估测误差(RPD)和均方根误差(RMSE)分别为0.77, 1.82和0.82。因此, 小波能量特征方法既能够提高数据的信噪比, 提升土壤有机质含量的估测精度, 又实现了土壤高光谱数据降维, 降低了模型复杂度, 可用于土壤有机质含量快速测定和土壤肥力动态监测等研究。

关键词: 土壤高光谱; 小波系数; 小波能量特征; 土壤有机质; 水稻土

中图分类号:S127 文献标志码:A

Determination of Soil Organic Matter Content Based on Hyperspectral Wavelet Energy Features

ZHANG Tao^1,^2,³, YU Lei^1,^2,^3,^*, YI Jun^1,^2,³, NIE Yan^1,^2,³, ZHOU Yong^1,^2,³

1. Hubei Provincial Key Laboratory for the Analysis and Simulation of Geographical Process, Central China Normal University, Wuhan 430079, China

2. College of Urban and Environmental Science, Central China Normal University, Wuhan 430079, China

3. Research Institute for Sustainable Development of Central China Normal University, Wuhan 430079, China

*Corresponding author

Abstract

There is no silver-bullet solution of eliminating noise during the acquisition process of soil hyperspectral. As the noise interference, the observations of soil spectra are in low signal-to-noise ratio, which affects the estimation accuracy of soil organic matter content. This paper attempts to adopt the wavelet energy features method to reduce the noise in soil hyperspectral and improve the estimation accuracy of soil organic matter content. The Yunlianghu Farm of Qianjiang City, Hubei Province, located in the hinterland of Jianghan Plain, was selected as the experimental area, and 80 samples of paddy soil with a depth of 0~20 cm were collectedin September 2016. After pretreatment (air drying, milling, sieving), soil sample spectral reflectance and determine soil organic matter contentwere collected in the laboratory. The concentration gradient method was employed to divide the whole sample set (80 samples) into a calibration set (54 samples) and a validation set (26 samples). Continuous wavelet transformation was performed using mexh as a wavelet basis function, transforming the soil hyperspectral into sensitive wavelet coefficients of 10 decomposition scales. Then the root mean square of the wavelet coefficients was calculated scale by scaleto define wavelet energy features, and the wavelet energy features vector was determined by the wavelet energy features. The correlation coefficients between the wavelet coefficients and the organic matter content were calculated scale by scale and wavelength by wavelength, and the wavelet coefficient which reaches the extremely significant level ( p<0.01) was defined as the sensitive wavelet coefficients. Principal component analysis was conducted to calculate the principal component loads of soil hyperspectral and wavelet energy features vector, respectively. The trend of principal component information of modeled independent variables before and after wavelet energy features transformation would be judged from the difference between the first principal component contribution rate and the spatial dispersion of the first three principal component scores degree. Moreover, regression models were established based on wavelet energy features vector and sensitive wavelet coefficients, respectively, to estimate soil organic matter content. The results showed that with the increase of soil organic matter content, the full-band reflectance decreased, but the spectral reflectance curves of different soil samples were similar, and the reflectance in the near-infrared bands (780~2 400 nm) was higher than that in the visible bands (400~780 nm). The sensitive wavelet coefficients corresponded to wavelengths of 494, 508, 672, 752, 1 838, and 2 302 nm. The first principal component contribution rates of soil hyperspectral and wavelet energy features vector were 96.28% and 99.13%, respectively. The first three principal component scatter points of wavelet energy features vector were more spatially aggregated than those of soil hyperspectral, which demonstrated that the wavelet energy features method effectively reduces the influence of noise. Comparing the estimation models of soil organic matter content, the multivariate linear regression model adopting wavelet energy features vector as the independent variable had the highest estimation accuracy, whose determination coefficients ( R²), relative estimate deviation (RPD), and the root mean squared error (RMSE) of validation set were 0.77, 1.82, and 0.82, respectively. Therefore, the wavelet energy features method which is proved to raise the signal-to-noise ratio of the data without adding to the complexity, could improve the estimation accuracy of soil organic matter and realize the dimensional reduction of soil hyperspectral data. This method can be applied to studies like on-the-go soil properties measurement and soil quality monitoring.

Keyword: Soil hyperspectral; Wavelet coefficients; Wavelet energy features; Soil organic matter; Paddy soil

文章图片

引言

土壤有机质是评价土壤质量的关键因素, 在调控农田肥力和作物生长方面发挥着重要作用^[1]。采用农化分析方法测定土壤有机质含量步骤繁琐、耗时耗力, 不利于土壤有机质的实时高效测定^[2]。土壤高光谱包含了丰富的土壤信息, 近年来, 被广泛应用于估测土壤理化性质(如土壤有机质、土壤水分及土壤全氮等), 取得了卓有成效的研究成果^{[3, 4]}。然而, 高光谱采集易受测试环境、样品质量和仪器本身等因素影响, 无法避免会产生噪声, 因此, 探索降噪方法对提升土壤有机质含量估测精度具有重要意义。

学者们尝试采用连续统去除, 一阶微分, 高通滤波等预处理方法, 消除样品背景及杂散光等因素所引起的噪声和增强特征信号, 但难以有效去除白噪声, 特别是随机和低频信号^{[5, 6]}。于雷等^[7]、廖钦洪等^[8]基于连续小波变换法将土壤高光谱分解成多尺度小波系数, 采用相关性分析法筛选与有机质含量具有极显著相关性的敏感小波系数(sensitive wavelet coefficients, SWC)建立估测模型; 蔡亮红等^[9]采用小波变换对土壤光谱进行分解, 通过竞争适应重加权采样法(competitive adaptive reweighted sampling, CARS)算法筛选SWC, 建立土壤水分高光谱估测模型, 这些研究表明小波变换方法能够克服白噪声影响, 将土壤高光谱分解为不同尺度的小波系数, 以更准确地局部描述和分离信号特征^{[10, 11, 12]}。但是, 通过提取SWC估测土壤属性, 难以充分利用不同尺度小波系数包含的有效光谱信息^[13]。 Pu等^[14]通过逐尺度提取全波段小波系数的能量特征(energy features, EF)组成小波能量特征向量(energy features vector, EF_V)估测森林叶面积指数; 宋开山等^[15]基于小波分析提取大豆叶片光谱的EF_V, 建立叶绿素a含量高光谱估测模型, 结果表明基于EF_V的回归模型可提高估测水平, 这对构建土壤有机质高光谱估测模型具有重要的参考价值。

本工作采用小波变换方法将土壤高光谱分解为10个尺度的小波系数, 利用均方根法计算各尺度的EF, 构建EF_V并建立基于EF_V的土壤有机质含量估测模型。目的是探索土壤高光谱噪声去除和信息提取方法, 以提高土壤有机质高光谱估测性能。

1 实验部分

1.1 土样的采集

试验区位于湖北省潜江市运粮湖管理区[图1(a)], 该区地势平坦, 水源充足, 第一产业较为发达, 是重要的粮油生产基地。试验区面积约为375 m× 300 m(11.25 ha)的连片田块, 土壤类型为水稻土, 是潜江市主要土类之一(占全市面积的55.22%); 常年采用中稻-油菜轮作的耕作制度。

2016年9月在试验区内均匀布设80个样点, 由北向南共5行(每行相距75 m)、由西向东共16列(每列相距25 m) [图1(b)]; 土样采集深度为0~20 cm, 用密封袋带回实验室经一系列预处理操作后用于后续分析。

	Figure Option View Download New Window
	图1 研究区(a)及采样方案(b)Fig.1 The study area (a) and sketch of sampling points (b)

1.2 土壤样本集的划分

将80个土样作为总体样本集, 根据土壤有机质含量从小到大排序, 采用浓度梯度法依次每隔两个样本取一个作为验证集, 共计26个样本, 其余54个样本作为建模集(表1)。

表1 样本统计特征 Table 1 Descriptive statistics of sample sets

1.3 土样光谱数据的测定

采用ASD FieldSpec3光谱仪测定土样光谱, 详细仪器参数、采集过程、参数设置及预处理可参见文献[16]。由于两端波长处的信噪比较低, 因此只保留400~2 400 nm。

1.4 小波变换和能量特征向量的计算

采用连续小波变换(CWT)对土壤高光谱数据进行转换, 通过小波分解后, 得到各尺度的小波系数^[17]

$W_{f} (a, b) \leq f; Ψ_{a, b} \geq \overset{+ \infty}{\int_{- \infty}} f (λ) Ψ_{a, b} (λ) d λ$ (1)

$Ψ_{a, b} = \frac{1}{\sqrt[]{a}} Ψ (\frac{λ - b}{a})$ (2)

其中, λ 为高光谱波段数, a为尺度参数, b为位移参数, f(λ )是光谱反射率, Ψ _a_,_b为基函数, W_f(a, b)为小波系数。 W_f(a, b)包括i和j两个维度表示为W_f(a_i, b_j), i为分解尺度, j为波段。

计算第i个分解尺度的小波系数的均方根, 作为该尺度的EF_i[式(3)]。

$E F_{i} = \sqrt[]{\frac{1}{j} \overset{j}{\sum_{j = 1}} W_{f} (a_{i}, b_{j})^{2}}$ (3)

构建EF_V[式(4)]。

$E F_{V} = [E F_{1}, E F_{2}, \dots, E F_{i}]$ (4)

1.5 敏感小波系数的确定

为对比EF_V与SWC模型估测效果, 采用相关性分析法筛选每一分解尺度与有机质含量相关性极显著(p< 0.01)的小波系数作为SWC[式(5)]。其中, R²为决定系数, X_sw为样本s在波长为w所对应的小波系数, Y_s为样本s的土壤有机质含量, ${\overset{̅}{X}}_{w}$ 为波长为w的小波系数平均值, $\overset{̅}{Y}$ 为土壤有机质含量的平均值。

$R^{2} = {(\frac{\overset{n}{\sum_{s = 1}} (X_{sw} - {\overset{̅}{X}}_{w}) (Y_{s} - \overset{̅}{Y})}{\sqrt[]{\overset{n}{\sum_{s = 1}} (X_{sw} {- \overset{̅}{X})}^{2}} \sqrt[]{\overset{n}{\sum_{s = 1}} (Y_{s} {- \overset{̅}{Y})}^{2}}})}^{2}$ (5)

1.6 模型构建与检验

偏最小二乘回归(PLSR)是最常用的高光谱估测模型之一, 其考虑了因变量的作用, 在处理高维数据、解决多重共线性时性能优越^[18]。多元线性回归(MLR)是一种较为简单的线性回归方法, 适用于样本数大于自变量数^[19]。分别以土壤高光谱(记为: Origin)、 SWC和EF_V为自变量, 有机质含量为因变量建立PLSR和MLR模型, 依次表示为Origin-PLSR, SWC-PLSR, EF_V-PLSR, SWC-MLR, EF_V-MLR。通过R², RMSE和RPD来检验估测模型精度^[20]。

2 结果与讨论

2.1 土壤光谱特征分析

不同土壤有机质含量的光谱反射率曲线(图2)基本特征相似, 可见光部分反射率低于近红外部分; 土样全波段反射率与有机质含量成反比。在可见光部分的土壤反射率曲线与近红外部分差异较大; 在1 400, 1 900和2 200 nm处, 出现具有不同吸收深度和宽度的吸收谷。

	Figure Option View Download New Window
	图2 土壤高光谱反射率Fig.2 Spectral reflectance curves of soil

2.2 敏感小波系数分析

以mexh为小波基函数, 计算土壤高光谱在10个尺度上(2^1-10)的小波系数, 分析小波系数与有机质含量的相关性, 得到R²分布状况(图3)。当p< 0.01, 则R²> 0.12, 由于波长众多, 难以提取敏感波长, 以图中红色部分(R²> 0.3)作为相关性极强的区域。不同分解尺度和波段下的R²存在明显差异, 说明经过小波变换, 与土壤有机质相关的细节特征信息被逐层释放, 每一层小波系数信息代表性各异。其中, 红色区域主要分布在490~500, 500~510, 670~680, 740~760, 1 830~1 840和2 300~2 310 nm, 分别对应尺度5, 4, 3, 3, 7和4, 对应最大值波长位置为494, 508, 672, 752, 1 838和2 302 nm, 将上述波长对应的小波系数确定为SWC。

	Figure Option View Download New Window
	图3 不同尺度R²分布Fig.3 R² distribution at different decomposition scales

2.3 小波能量特征向量分析

根据EF计算公式, 每个尺度提取一个EF, 最终每个样本形成10维EF_V。主成分分析法(PCA)可有效降低数据维度, 以少数几个主成分变量反映原始信息。在本工作中, PCA用于分析土壤高光谱和EF_V的第一主成分贡献率及前三个主成分得分的空间离散程度, 判断EF转换前后变量的主成分信息变化趋势。 EF_V的第一主成分贡献率为99.13%, 比原始光谱的第一主成分贡献率高出3%[图4(a)], 接近于100%, 表明EF_V第一主成分所包含的信息量较土壤高光谱有所增加, 且具有极强的信息代表性。前三个主成分得分所建立的三维散点图[图4(b)]更为形象地表明EF_V的降维效果, 原始光谱各样本主成分得分散点分布广泛, 样本间差异显著, 主成分所含信息分散; EF_V各主成分得分散点十分聚集, 表明噪声消除后样本间差异降低。通过PCA分析, EF_V在降低光谱数据维度、消减光谱噪声的同时提高了样本变量信息贡献率。

	Figure Option View Download New Window
	图4 土壤高光谱、能量特征向量前三个主成分贡献率(a)及得分散点(b)Fig.4 The first three principal component contribution rate (a) and the score scatter point (b) of soil original spectral and the root mean square energy feartures vector

2.4 估测模型比较

由表2可以看出, 同一自变量集的MLR模型估测效果要优于PLSR模型, 表明当变量数小于样本数时(在此SWC和EF_V自变量数分别为6和10个), MLR模型具有更好的适用性, 可更为准确地表达土壤有机质同高光谱间的关系, 简化模型结构。由此表明, 在选取估测模型时, 应充分考虑变量及样本特征。

基于不同自变量集的最优模型分别为Origin-PLSR, SWC-MLR, EF_V-MLR模型, 对应R²依次为0.70, 0.73和0.77, 表明SWC-MLR, EF_V-MLR模型估测效果要优于Origin-PLSR模型。在所有估测模型中EF_V-MLR模型表现最优, RPD达到1.82, 可较为准确地估测土壤有机质含量。利用小波分析可以降低噪声干扰, 放大土壤有机质敏感信息。

通过提取SWC和EF_V均可降低数据维度, 减少模型复杂度, 提升模型估测性能。基于EF_V的模型要优于SWC模型, 表明EF在小波系数的基础上, 在每个尺度上计算小波分解系数的能量, 形成特征向量, 相比于代表部分敏感信息的SWC, EF_V更为综合地表征了光谱信息。同时, EF无需事先获取待测属性且数据压缩维度可控, 相比于基于相关性分析提取SWC, 该方法更具推广性和可重复性, 可尝试用于多领域高光谱数据提取。

表2 不同估测模型的效果比较 Table 2 Comparison of different estimation model results

3 结论

基于小波变换方法提取EF, 有效减少了土壤光谱噪声影响, 提升了土壤有机质高光谱估测精度。研究结果表明, 小波变换可放大光谱细节特征, 降低噪声干扰, SWC可代表部分与土壤有机质相关性较强的光谱信息; EF充分利用了各尺度小波系数包含的光谱信息, 提高了光谱数据信噪比, 具有极佳的信息代表性; 在所有模型中, EF_V-MLR模型表现最优, 可有效降低数据维度与模型复杂度, 对于快速估测田间土壤属性, 推动农业精准化管理具有一定的参考价值。但EF代表的是不同尺度小波系数整体信息, 如何更加准确地结合不同尺度的局部特征, 值得深入研究。

参考文献

文献列表

[1]	Dotto A C, Dalmolin R S D, Caten A T, et al. Geoderma, 2018, 314: 262. [本文引用:1]
[2]	ZHOU Qian-qian, DING Jian-li(周倩倩, 丁建丽). Acta Pedologica Sinica(土壤学报), 2018, 55(2): 313. [本文引用:1]
[3]	Rossel R A V, Behrens T, Ben-Dor E, et al. Earth-Science Reviews, 2016, 155: 198. [本文引用:1]
[4]	Nouri M, Gomez C, Roger J M, et al. Geoderma, 2017, 298: 54. [本文引用:1]
[5]	ZHANG Rui, LI Zhao-fu, PAN Jian-jun(张锐, 李兆富, 潘剑君). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2017, 33(1): 175. [本文引用:1]
[6]	CHEN Hong-yan, ZHAO Geng-xing, LI Xi-can, et al(陈红艳, 赵庚星, 李希灿, 等). Scientia Agricultura Sinica(中国农业科学), 2012, 45(7): 1425. [本文引用:1]
[7]	YU Lei, HONG Yong-sheng, ZHOU Yong, et al(于雷, 洪永胜, 周勇, 等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2016, 36(5): 1428. [本文引用:1]
[8]	LIAO Qin-hong, GU Xiao-he, LI Cun-jun, et al(廖钦洪, 顾晓鹤, 李存军, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2012, 28(23): 132. [本文引用:1]
[9]	CAI Liang-hong, DING Jian-li(蔡亮红, 丁建丽). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2017, 33(16): 144. [本文引用:1]
[10]	Wang H F, Huo Z G, Zhou G S, et al. Plant Physiology & Biochemistry, 2016, 98: 39. [本文引用:1]
[11]	CHEN Hong-yan, ZHAO Geng-xing, LI Xi-can, et al(陈红艳, 赵庚星, 李希灿, 等). Chinese Journal of Applied Ecology(应用生态学报), 2011, 22(11): 2935. [本文引用:1]
[12]	Lark R M. Geoderma, 2016, 266: 84. [本文引用:1]
[13]	Yun Y H, Wang W T, Tan M L, et al. Analytica Chimica Acta, 2014, 807(1): 36. [本文引用:1]
[14]	Pu R, Gong P. Remote Sensing of Environment, 2004, 91(2): 212. [本文引用:1]
[15]	SONG Kai-shan, ZHANG Bai, WANG Zong-ming, et al(宋开山, 张柏, 王宗明, 等). Chinese Journal of Plant Ecology(植物生态学报), 2008, 32(1): 152. [本文引用:1]
[16]	ZHU Ya-xing, YU Lei, HONG Yong-sheng, et al(朱亚星, 于雷, 洪永胜, 等). Scientia Agricultura Sinica(中国农业科学), 2017, 50(22): 4325. [本文引用:1]
[17]	Bruce L M, Li J. Geoscience & Remote Sensing IEEE Transactions on, 2001, 39(7): 1540. [本文引用:1]
[18]	Yu X, Liu Q, Wang Y B, et al. Catena, 2016, 137: 340. [本文引用:1]
[19]	LI Pei-zhe(李培哲). Statistics and Decision(统计与决策), 2012, (24): 89. [本文引用:1]
[20]	Xu S X, Shi X Z, Wang M Y, et al. Geoderma, 2016, 267: 174. [本文引用:1]

2018

0.0

... 引言土壤有机质是评价土壤质量的关键因素, 在调控农田肥力和作物生长方面发挥着重要作用^[1] ...

2018

0.0

... 采用农化分析方法测定土壤有机质含量步骤繁琐、耗时耗力, 不利于土壤有机质的实时高效测定^[2] ...

2016

0.0

... 土壤高光谱包含了丰富的土壤信息, 近年来, 被广泛应用于估测土壤理化性质(如土壤有机质、土壤水分及土壤全氮等), 取得了卓有成效的研究成果^[3,4] ...

2017

0.0

2017

0.0

ZHANG

Rui

, LI

Zhao-fu

, PAN

Jian-jun

(张锐, 李兆富, 潘剑君). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2017, 33(1): 175.

高光谱遥感可以实现水稻土排水期有机碳含量的快速预测，但土壤反射率受多种噪声的影响，有机碳光谱信号探测受阻，预测模型性能低下，如何在去除噪声的同时最大限度地保持有机碳光谱信号十分重要。以原状新鲜水稻土为研究对象，采用Bior1.3小波系对反射光谱进行1～7层小波包变换，通过相关分析确定最大分解层；将原始反射率至最大分解层以内的各层光谱相关系数组成相关系数集，采用局部最相关算法（local correlation maximization，LCM）构造土壤有机碳最优光谱；最后基于最优光谱建立有机碳含量偏最小二乘预测模型并进行分析。结果显示：1）随着小波包分解层数的增加，土壤反射率与有机碳含量的相关性不断增强，到第6层达到最高，确定为小波包最大分解层；2）基于LCM构造的最优光谱比未去噪光谱平滑，比小波包去噪光谱保留了更多光谱细节；3）未去噪光谱、小波包去噪光谱和LCM最优光谱有机碳预测模型的验证决定系数分别为0.693、0.727和0.781，均方根误差为1.952、1.840和1.679 g/kg，残留预测偏差为1.85、1.97和2.17。小波包-局部最相关算法在去噪同时有效保持了土壤有机碳光谱信号，可提高水稻土有机碳含量高光谱预测精度。

... 学者们尝试采用连续统去除, 一阶微分, 高通滤波等预处理方法, 消除样品背景及杂散光等因素所引起的噪声和增强特征信号, 但难以有效去除白噪声, 特别是随机和低频信号^[5,6] ...

2012

0.0

CHEN

Hong-yan

, ZHAO

Geng-xing

, LI

Xi-can

, et al(陈红艳, 赵庚星, 李希灿, 等). Scientia Agricultura Sinica(中国农业科学), 2012, 45(7): 1425.

【Objective】 This study aimed at improving the precision and practicability of the soil available potassium estimation model by removing the noise of soil hyperspectral reflectance. 【Method】 Seventy-six soil samples with similar soil organic matter, nitrogen and phosphorus element contents and different potassium element contents were selected. The first derivative spectrum of the soil sample logarithmic reflectance was decomposed to multiple levels by using four kinds of wavelet function, respectively. The low frequency wavelet coefficients were obtained, and the hyperspectral estimation models of soil available potassium content were built. 【Result】The results showed that the low frequency wavelet coefficients of 1-3 levels could represent the original spectrum. Based on the low frequency coefficients of different wavelet functions at the same level, the precise of soil available potassium estimation model showed a little difference. The model built with the low frequency coefficient of the second decomposition level using the Bior 1.3 function had comparatively high accuracy and was chosen as the best model. With the data reducing to 25% and reflecting 95.6% information of the input spectrum, the model building R2 reached 0.976 and RMSE was 10.66 mg•kg-1, which was validated to have fairly good forecast accuracy. 【Conclusion】Therefore, wavelet analysis for obtaining wavelet coefficients can not only extract the soil hyperspectral information, but also compress data, which is feasible to forecast soil potassium content in combination with partial least squares regression method.

1.山东农业大学资源与环境学院，山东泰安 271018 2.山东农业大学信息科学与工程学院，山东泰安 271018 3.山东农业大学园艺科学与工程学院，山东泰安 271018 4.山东省沾化县国土资源局，山东沾化 256800

【目的】对土壤高光谱数据去噪提纯，提高土壤速效钾含量高光谱估测模型的精度和实用性。【方法】选取土壤有机质、碱解氮、有效磷含量近似而速效钾含量差异较大的样本76个，对土壤样本反射率对数的一阶导数光谱分别基于4种函数进行多层小波离散分解；提取小波低频系数，构建土壤速效钾含量高光谱估测模型。【结果】小波分解1—3层获得的低频系数可用以代表原始光谱。基于各小波函数相同尺度的低频系数，土壤速效钾含量估测建模精度差异不大。其中基于Bior 1.3函数分解的第2层低频系数建模精度略高，作为最佳估测模型，在数据压缩到25%、反映输入光谱信息95.6%的基础上，建模R2达到0.976，RMSE为10.66 mg•kg-1，经验证模型具有较好的预测准确度。【结论】通过小波分析获得小波系数，既提取了土壤高光谱信息，又对数据进行了压缩，结合偏最小二乘回归预测土壤速效钾含量是可行的。

2016

0.0

... 于雷等^[7]、廖钦洪等^[8]基于连续小波变换法将土壤高光谱分解成多尺度小波系数, 采用相关性分析法筛选与有机质含量具有极显著相关性的敏感小波系数(sensitive wavelet coefficients, SWC)建立估测模型 ...

2012

0.0

LIAO

Qin-hong

, GU

Xiao-he

, LI

Cun-jun

, et al(廖钦洪, 顾晓鹤, 李存军, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2012, 28(23): 132.

Hyperspectral remote sensing technology had been widely used in the estimation of soil organic matter due to its non-destructive, rapid, and high spectral resolution characteristics. The 64 fluvo-aquic soil organic matter (SOM) obtained from Beijing Shunyi district had been estimated successfully by using the hyperspectral reflectance based on continuous wavelet transform (CWT), the results had also been compared with four common spectral transformation methods, it showed that the fluvo-aquic soil had the same spectral curves with other types soils, and the absorption peaks appeared in the visible and near infrared bands after the spectral curves removed by hull curve. The sensitive bands for estimating the SOM were 1194nm, 486nm, and 866nm, and the corresponding wavelet decomposition scales were 2, 3, and 4. The R2 of multiple linear regression model built between the wavelet energy coefficients and SOM was 0.67 by using the CWT, the R2and RMSE between the measured value and predicted value were 0.75 and 0.21, respectively, while the highest R2 of estimation model built by the four common spectral transformation methods was 0.09, which showed that the CWT is more suitable for estimating the fluvo-aquic soil organic matter content. Through the interpolation analysis of Kringing, the more sampling points should be increased in the southeast of Shunyi district for improving the precision of models.

土壤有机质含量快速估算对于土壤肥力评价、土壤信息化管理和精准施肥具有重要意义。该文通过对北京顺义地区64个土壤样品高光谱曲线进行连续小波变换，估算了该地区潮土有机质质量分数，并与4种常用光谱变换方法进行了比较。结果表明，潮土具有与其他类型土壤类似的光谱曲线，经过去包络线处理后，在可见与近红波段都出现了明显吸收峰；采用连续小波变换方法所确定的潮土有机质估算的敏感波段为1 194、486和866 nm，对应小波分解尺度为2，3和4；利用小波能量系数与有机质质量分数所构建的多元线性回归模型的决定系数R2为0.67，模型实测值与预测值的检验精度R2为0.75，RMSE为0.21；而采用4种常用光谱变换方法建立的潮土有机质估测模型的R2最高只有0.09，说明连续小波变换方法更适合于潮土有机质质量分数估测。Kringing插值分析表明，应在顺义地区东南部增加取样点，以提高模型估算精度。该研究可为潮土土壤肥力的快速测定提供参考。

2017

0.0

... 蔡亮红等^[9]采用小波变换对土壤光谱进行分解, 通过竞争适应重加权采样法(competitive adaptive reweighted sampling, CARS)算法筛选SWC, 建立土壤水分高光谱估测模型, 这些研究表明小波变换方法能够克服白噪声影响, 将土壤高光谱分解为不同尺度的小波系数, 以更准确地局部描述和分离信号特征^[10,11,12] ...

2016

0.0

2011

0.0

CHEN

Hong-yan

, ZHAO

Geng-xing

, LI

Xi-can

, et al(陈红艳, 赵庚星, 李希灿, 等). Chinese Journal of Applied Ecology(应用生态学报), 2011, 22(11): 2935.

A total of 60 soil samples with approximate contents of N, P, and K and greatly different content of organic matter were selected by statistical analysis. Through hyper-spectral detection and analysis, the first derivative spectrum of the soil logarithmic reflectance was obtained, and was decomposed by the Bior 1.3 wavelet function. The approximative signal of the lowest frequency and the noise signal of the highest frequency were removed from the input spectrum so as to obtain the characteristic spectrum corresponding to soil physical and chemical parameters. The sensitive bands of soil organic matter were selected by correlation analysis, and the forecasting models were built by multiple regression analysis, based on the sensitive bands and the characteristic spectrum, respectively. Through comparison analysis, the optimal wavelet decomposing resolution for extracting the characteristic spectrum of soil organic matter was ascertained, and the best forecasting model was established. The best wavelet decomposing resolution was 9, followed by 8 and 10. Based on the characteristic spectrum of wavelet de composing of 9 resolutions, the model R 2 reached 0.89, which was increased by 0.31 as compared to the model based on sensitive bands, and increased by 0.10 as compared to the model based on the original spectrum.

1College of Resources and Environment, Shandong Agricultural University, Tai’an 271018, Shandong, China;2College of Information Science and Engineering, Shandong Agricultural University, Tai’an 271018, Shandong, China; 3Zhanhua Land Resources Bureau, Zhanhua 256800, Shandong, China; 4Hekou Branch of Dongying Land Resources Bureau, Hekou 257200, Shandong, China

利用统计分析方法选取了土壤N、P、K元素含量近似而有机质含量差异较大的样本60个，通过高光谱探测分析获得样本反射率对数的一阶导数光谱，采用Bior 1.3函数进行多层离散小波分解，剔除低频近似信号和高频噪声信号，得到反映土壤理化参数的特征光谱曲线；采用相关分析筛选土壤有机质含量的显著相关波段，基于显著相关波段和特征光谱曲线分别构建土壤有机质含量高光谱多元回归估测模型；通过比较分析，确定了提取土壤有机质特征光谱的最佳小波分解尺度并构建了最佳预测模型.结果表明：提取土壤有机质特征光谱的最佳小波分解层数是9层，其次是8层和10层；基于小波9层分解特征光谱曲线的有机质含量估测模型最佳，其决定系数（ R 2 ）为0.89，比基于显著相关波段构建模型的 R 2 增加了0.31，比基于原始光谱所构建模型的 R 2 增加了0.10.

2016

0.0

2014

0.0

... 但是, 通过提取SWC估测土壤属性, 难以充分利用不同尺度小波系数包含的有效光谱信息^[13] ...

2004

0.0

... Pu等^[14]通过逐尺度提取全波段小波系数的能量特征(energy features, EF)组成小波能量特征向量(energy features vector, EF_V)估测森林叶面积指数 ...

2008

0.0

SONG

Kai-shan

, ZHANG

Bai

, WANG

Zong-ming

, et al(宋开山, 张柏, 王宗明, 等). Chinese Journal of Plant Ecology(植物生态学报), 2008, 32(1): 152.

Aims A growing number of studies have focused on evaluating spectral indices in terms of their sensitivity to vegetation biophysical parameters. We use a regression model, based on wavelet-transformed reflectance, and vegetation indices ( VI ) to estimate a wide range of soybean ( Glycine max ) canopy reflectances to study the sensitivity of wavelet-transformed reflectance and vegetation in dices to soybean chlorophyll a concentration. We modify some VI to enhance their sensitivity to variations in chlorophyll a concentration.  Methods We collected soybean canopy hyperspectral reflectance and chlorophyll a concentration data in 2003 and 2004 at two sites in the black soil belt of China. We correlated reflectance, derivative reflectance and soybean chlorophyll a concentration and regressed vegetation indices ( NDVI , SAVI, RDVI and MSRI ) and soybean chlorophyll a concentration. We transformed soybean canopy reflectance with wavelet analysis and applied extracted wavelet energy coefficient in a regression model for estimation of chlorophyll a concentration. Important findings Soybean canopy reflectance shows a negative correlation with chlorophyll a concentration in the visible spectral region, while it shows a positive correlation with soybean chlorophyll a concentration in the near-infrared region. Reflectance derivative has a strong relationship with soybean chlorophyll a concentration in the blue, green and red edge spectral region, with maximum correlation coefficient in the red-edge region. Four vegetation indices have strong correlations with soybean chlorophyll a concentration, with R 2 >0.75. The s ingle variable regression model based upon wavelet-extracted reflectance energy can accurately estimate soybean chlorophyll a concentration, with R 2 about 0.75, while R 2 was 0.85 with the multivariate regression model. Our study indicated that wavelet analysis can be applied to in situ collected hyperspectral data for soybean chlorophyll a concentration estimation with accurate prediction and in the future wavelet analysis methods should be applied to hyperspectral data for estimation of other vegetation biophysical and biochemical parameters.

1Northeast Institute of Geography and Agricultural Ecology, Chinese Academy of Sciences, Changchun 130012, China; 2 Graduate University of Chinese Academy of Sciences, Beijing 100049, China

2003和2004年分别在长春市良种场和中国科学院海伦黑土生态实验站实测了大田耕作与水肥耦合作用下大豆（ Glycine max ）冠层高光谱反射率与叶绿素a含量数据，对光谱反射率、微分光谱与叶绿素a含量进行了相关分析；采用归一化植被指数(Normalized diffe rence vegetation index, NDVI )、土壤调和植被指数(Soil-adjusted vegetation index, SAVI )、再归一植被指数(Renormalized difference vegetation index, RDVI )、第二修正比值植被指数(Modified second ratio index, MSRI )等建立了大豆叶绿素a反演模型；应用小波分析对采集的光谱反射率数据进行了能量系数提取，并以小波能量系数作为自变量进行了单变量与多变量回归分析，对大豆叶绿素a进行了估算。研究结果表明，大豆叶绿素a 与可见光光谱反射率相关性较好，并在红光波段取得最大值( R 2 >0.70)，但在红边处，微分光谱与大豆叶绿素a的相关性较反射率好得多，在其它波段则相反；由 NDVI 、 SAVI 、 RDVI 、 MSRI 等植被指数建立的估算模型可以提高大豆叶绿素a的估算精度（ R 2 >0.75）；小波能量系数回归模型可以进一步提高大豆叶绿素a含量的估算水平，以一个特定小波能量系数作为自变量的回归模型，大豆叶绿素a回归决定系数 R 2 高达 0.78；多变量回归分析结果表明，大豆叶绿素a实测值与预测值的线性回归决定系数 R 2 均高达0.85。以上结果表明，小波分析可以对高光谱进行特征变量提取，并可在一定程度上提高大豆生理参数反演精度。

... 宋开山等^[15]基于小波分析提取大豆叶片光谱的EF_V, 建立叶绿素a含量高光谱估测模型, 结果表明基于EF_V的回归模型可提高估测水平, 这对构建土壤有机质高光谱估测模型具有重要的参考价值 ...

2017

0.0

ZHU

Ya-xing

, YU

Lei

, HONG

Yong-sheng

, et al(朱亚星, 于雷, 洪永胜, 等). Scientia Agricultura Sinica(中国农业科学), 2017, 50(22): 4325.

【Objective】The objective of this study is to explore the hyperspectral features and response regularity of the soil organic matter, and to select the sensitive wavelengths of soil organic matter, so as to reduce complexity of hyperspectral estimation model of soil organic matter and improve robustness of the model, which is to provide theoretical support to quantitatively monitor the soil fertility of farmland by using the hyperspectral technology. 【Method】 A total of 130 fluvo-aquic soil samples were collected from Jianghan plain, of which 40 were the training set samples. The soil organic matter content (SOMC raw ) and spectral reflectance (SR raw ) were measured from total samples, and an experiment of removal of organic matter was performed using the training set samples, and then we measured the soil organic matter content (SOMC rem ) and spectral reflectance (SR rem ) from samples of removal of organic matter. By calculating the difference and rate of change between SR raw and SR rem from training set samples, we could analyze how the content changes of soil organic matter itself influence the spectral features. The soil organic matter sensitive wavelengths were determined by the methods of uninformative variables elimination (UVE) and competitive adaptive reweighted sampling (CARS). The calibration set with 45 samples was utilized to build the soil organic matter estimation models base on partial least squares regression (PLSR) and back propagation neural network (BPNN), and the validation set of 45 samples was utilized to test whether sensitive wavelengths were suitable for the same type soil. 【Result】 The experiment of removal of organic matter showed that the average spectral reflectance of test soil samples increased at full-spectrum with removing organic matter content, especially at the visible spectrum; after the comparison of UVE, CARS, UVE-CARS, and CARS-UVE, the optimal method of variables selection was UVE-CARS. The method of UVE-CARS provided 84 selected variables which were the soil organic matter sensitive wavelengths with coverage area of 561-721, 1 920-2 280 nm. Based on soil organic matter sensitive wavelengths, the PLSR and BPNN had better performance than full-spectrum model, and BPNN was better than PLSR in predictive ability with its value of R 2 , RMSE, RPD, MAE, MRE were 0.74, 1.33 g·kg -1 , 2.02, 1.04 g·kg -1 , 6.2%, respectively, so it could effectively estimate soil organic matter. 【Conclusion】 The soil organic matter sensitive wavelengths from training set could effectively estimate soil organic matter content in this test area with the same type samples. In addition, modeling of sensitive wavelengths by obtaining from the experiment of removal of organic matter and variables selection method could not only compress input wavelengths down into 4.2% of full-spectrum, but also enhance the estimation accuracy and reduce the model complexity. In this study, it provided a new approach to quickly and accurately evaluate soil organic matter content in the farmland.

Key Laboratory for Geographical Process Analysis and Simulation, Hubei Province, Central China Normal University/ College of Urban & Environmental Science, Central China Normal University, Wuhan 430079

【目的】探究土壤有机质的高光谱特征及响应规律,优选土壤有机质的敏感波长,降低土壤有机质高光谱估测模型复杂度,提高模型稳健性,为利用高光谱技术对农田土壤肥力的定量监测提供理论支撑。【方法】采集江汉平原潮土土样 130个,将其中40个样本作为训练集,测量其去有机质前、后的土壤有机质含量及光谱数据,计算差值及变化率,分析土壤有机质含量变化对光谱特征的影响,结合无信息变量消除（uninformative variables elimination,UVE）、竞争适应重加权采样（competitive adaptive reweighted sampling,CARS）变量优选方法确定土壤有机质敏感波长；采用45个建模集样本,基于偏最小二乘回归(partial Least Squares Regression,PLSR)和反向传播神经网络（ back propagation neural network, BPNN）建立土壤有机质含量的估算模型；利用 45个验证集样本检验敏感波长对同类土壤的适用性。【结果】通过有机质去除试验,供试土壤的平均光谱反射率在全波段均有所增加,在可见光波段变化率高于近红外波段;比较UVE、CARS、UVE-CARS、CARS-UVE这4种变量优选方法,得到最佳变量优选方法为UVE-CARS,该方法从2001个波长变量中优选得到84个变量作为土壤有机质的敏感波长,分布于 561— 721、1 920—2 280 nm波段覆盖范围;基于敏感波长的PLSR、BPNN模型性能均优于全波段模型,其中,基于敏感波长的BPNN模型的估测能力高于PLSR,模型验证集 R 2 、RMSE、RPD、MAE、MRE值分别为0.74、1.33 g ·kg -1 、 2.02、1.04 g ·kg -1 、 6.2%,可实现土壤有机质含量的有效估测。【结论】通过训练集获得的土壤有机质敏感波长,能够实现对该试验区同种土壤类型样本土壤有机质含量的有效估测；利用去有机质试验结合变量优选方法确定的敏感波长建模,不仅将输入波长压缩至全波段波长数目的 4.2 %,而且提升了模型估测精度,降低了变量维度和模型复杂度,为快速准确评估农田土壤有机质含量提供了新途径。

... 3 土样光谱数据的测定采用ASD FieldSpec3光谱仪测定土样光谱, 详细仪器参数、采集过程、参数设置及预处理可参见文献[16] ...

2001

0.0

... 4 小波变换和能量特征向量的计算采用连续小波变换(CWT)对土壤高光谱数据进行转换, 通过小波分解后, 得到各尺度的小波系数^[17] ...

2016

0.0

... 6 模型构建与检验偏最小二乘回归(PLSR)是最常用的高光谱估测模型之一, 其考虑了因变量的作用, 在处理高维数据、解决多重共线性时性能优越^[18] ...

2012

0.0

... 多元线性回归(MLR)是一种较为简单的线性回归方法, 适用于样本数大于自变量数^[19] ...

2016

0.0

... 通过R², RMSE和RPD来检验估测模型精度^[20] ...