光谱学与光谱分析, 2019, 39 (11): 3553, 网络出版: 2019-12-02  

倒伏胁迫下的玉米冠层结构特征变化与光谱响应解析

Structural Characteristics Change and Spectral Response Analysis of Maize Canopy under Lodging Stress
作者单位
1 农业部农业遥感机理与定量遥感重点实验室, 北京农业信息技术研究中心, 北京 100097
2 国家农业信息化工程技术研究中心, 北京 100097
3 北京市农业物联网工程技术研究中心, 北京 100097
4 山东科技大学测绘科学与技术学院, 山东 青岛 266590
5 北华航天工业学院计算机与遥感信息技术学院, 河北 廊坊 065000
摘要
倒伏胁迫是玉米生产中的主要灾害之一, 严重影响玉米的产量、 品质和机械收获能力。 解析不同倒伏胁迫强度下玉米冠层结构变化规律及其光谱响应机理, 是玉米倒伏灾情大范围遥感监测的基础。 分别在玉米抽雄期、 灌浆中期设置茎倒、 茎折、 根倒3种强度的倒伏处理, 基于田间多频次持续观测实验, 分析生育期、 倒伏类型对玉米冠层结构动态变化及其自我恢复能力的影响; 采用传统光谱变换与连续小波变换方法对倒伏玉米冠层高光谱进行处理, 选取叶面积密度(LAD)为玉米倒伏冠层结构特征指标, 筛选叶面积密度最佳敏感波段和小波系数, 基于随机森林法构建叶面积密度高光谱响应模型, 利用未参与建模的实测样本验证模型精度, 重点探讨小波分解尺度和光谱分辨率对LAD光谱响应能力的影响规律。 研究结果表明: 叶面积密度作为单位体积内叶面积总量的冠层结构表征指标, 与倒伏胁迫强度具有较好的响应关系, 灌浆期的倒伏玉米LAD普遍高于抽雄期, 抽雄期LAD整体表现为茎折>根倒>茎倒>未倒伏, 灌浆期LAD整体表现为根倒>茎折>茎倒>未倒伏; 经连续小波变换后, 玉米倒伏冠层光谱对玉米倒伏LAD的响应能力普遍优于传统光谱变换, 随着小波分解尺度的增加, LAD与敏感波段的相关性越强, 其中10尺度相关系数最高, 达0.74; 连续小波变换所构建的模型精度普遍优于传统光谱变换, 其中由原始光谱小波变换后构建的LAD响应模型精度最高, 检验样本的R2为0.811, RMSE为1.763, 表明连续小波变换技术可凸显和利用冠层光谱中的细微信息。 因此, 叶面积密度可有效定量表征不同倒伏胁迫程度的玉米冠层结构变化特征, 连续小波变换能有效提升冠层光谱对倒伏玉米结构变化的响应能力, 基于随机森林法构建的倒伏玉米叶面积密度诊断模型具有较高的精度和稳定性, 可为区域尺度的夏玉米倒伏灾情遥感监测提供先验知识。
Abstract
Lodging stress is one of the main disasters in maize production, which seriously affects the yield and quality of maize and mechanical harvesting ability. It is the basis of remote sensing monitoring of maize large-scale lodging disasters to analyze the changes of maize canopy structure and spectral response characteristics under different lodging stress. Stem lodging, stem fold and root lodging were set up in the tasseling stage and the middle filling period. Based on the field continuous observation experiments, the effects of growth stages and lodging types on dynamic changes of canopy structure and self-recovery ability of maize were analyzed. Hyperspectral data of lodging maize canopy were processed by the traditional spectral transformation and continuous wavelet transform. Leaf area density (LAD) was taken as the index of the lodging maize canopy structure characteristics. The best sensitive bands and wavelet coefficients of leaf area density were selected. The hyperspectral response model of leaf area density was constructed based on random forest method, andthe accuracy of the model was verified by the measured samples which were not involvedin modeling. Focus on the influence of wavelet decomposition scale and spectral resolution on LAD spectral response ability. The results showed that leaf area density, as a canopy structure indicator of total leaf area per unit volume, had a good response relationship with lodging stress intensity. The LAD of lodging maize in the filling stage was generally higher than that in the tasseling stage. The LAD of the tasseling stage was as follows: stem fold>root lodging>stem lodging>no lodging. The LAD of the filling stage was as follows: root lodging>stem fold>stem lodging>no lodging. After continuous wavelet transform, the response ability of maize lodging canopy spectrum to leaf area density is generally better than that of traditional spectral transform. The response ability of lodging maize canopy spectrum to leaf area density after continuous wavelet transform is generally better than that of traditional spectral transform. As the wavelet decomposition scale increases, the correlation between LAD and sensitive bands is stronger, and the correlation coefficient of 10 scale is the highest, reaching 0.74. The accuracy of the model constructed by continuous wavelet transform is generally better than that of traditional spectral transform. The model constructed by the original spectral wavelet transform has the highest precision, the R2 of test sample is 0.811, and the RMSE is 1.763. It showed that continuous wavelet transform technology can highlight and utilize the subtle information in the canopy spectra. Therefore, leaf area density can effectively quantify the variation characteristics ofmaize canopy structure under different lodging stress. Continuous wavelet transform can effectively improve the response of the canopy spectrum to the structural parameters of the lodging maize. The model of lodging maize leaf area density based on random forest method has high accuracy and stability, which can provide prior knowledge for remote sensing monitoring of summer maize lodging disaster at regional scale.

束美艳, 顾晓鹤, 孙林, 朱金山, 杨贵军, 王延仓, 孙乾, 周龙飞. 倒伏胁迫下的玉米冠层结构特征变化与光谱响应解析[J]. 光谱学与光谱分析, 2019, 39(11): 3553. SHU Mei-yan, GU Xiao-he, SUN Lin, ZHU Jin-shan, YANG Gui-jun, WANG Yan-cang, SUN Qian, ZHOU Long-fei. Structural Characteristics Change and Spectral Response Analysis of Maize Canopy under Lodging Stress[J]. Spectroscopy and Spectral Analysis, 2019, 39(11): 3553.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!