摘要
植株氮含量(PNC)是评价作物长势和营养状况的重要指标,快速准确获取作物的PNC信息可为农田管理策略的制定与实施提供重要依据。已有研究表明,仅采用影像的光谱信息估算作物的PNC存在饱和现象,该研究尝试采用植被指数(VIs)结合二维离散小波分解技术(DWT)提取的多个尺度的高频信息(HFI)构建一种光谱空间特征(VIs+HFI),探究VIs、 HFI和VIs+HFI估算PNC的能力。首先,以无人机为遥感平台获取马铃薯现蕾期、块茎形成期、块茎增长期、淀粉积累期和成熟期5个氮营养关键生育期的数码影像并实测各生育期的PNC数据。其次,基于预处理的无人机影像,提取各生育期冠层的光谱信息构建VIs,并采用DWT提取各生育期1~5尺度的HFI。然后,将各生育期提取的VIs和HFI与马铃薯PNC进行相关性分析,分别筛选出相关系数绝对值较大的前7个VIs和前10个HFI。为降低共线性对实验结果的影响,根据KMO检验结果对筛选的HFI进行主成分分析(PCA)降维处理。最后,采用岭回归和极限学习机(ELM) 2种方法分别以VIs、 HFI主成分和VIs+HFI主成分为模型变量构建马铃薯各生育期的PNC估算模型,并进行评估。结果表明:(1)马铃薯各生育期,1~5尺度的HFI对估算PNC均有贡献。(2)以VIs+HFI为模型变量构建的马铃薯PNC估算模型的精度和稳定性高于单一VIs和HFI。(3)马铃薯各生育期,以岭回归方法构建的PNC估算模型优于ELM方法。其中,以VIs+HFI为模型变量构建的PNC估算模型效果最优,5个生育期的建模R^(2)分别为0.833、 0.764、 0.791、 0.664和0.435, RMSE分别为0.332%、 0.297%、 0.275%、 0.286%和0.396%;NRMSE分别为9.113%、 9.425%、 10.336%、 9.547%和15.166%,该研究可为马铃薯氮营养状况的实时高效监测提供一种新的技术支撑。
Plant nitrogen content(PNC)is essential for evaluating crop growth and nutritional status.Obtaining crop PNC information quickly and accurately can provide an important basis for formulating and implementing farmland management strategies.Existing studies have shown saturation in estimating crop PNC using only the spectral information of images.Therefore,this research attempted to use vegetation indices(VIs)combined with two-dimensional discrete wavelet decomposition technology(DWT)to extract high-frequency information(HFI)at multiple scales.It was constructing a spectral,spatial feature(VIs+HFI)and exploring the ability of VIs,HFI,and VIs+HFI to estimate PNC.First,the UAV was a remote sensing platform to obtain digital images of the five critical nitrogennutrient growth periods of potato budding,tuber formation,tuber growth,starch accumulation,and maturity.It measured PNC data for each growth period.Secondly,based on the pre-processed UAV images,the spectral information of the canopy of each growth period was extracted to construct VIs,and the DWT was used to extract the HFI of each growth period 1~5 scales.Then,the VIs and HFI extracted from each growth period were correlated with the ground-truthed PNC data.The top 7 VIs and the top 10 HFI with larger absolute correlation coefficient values were screened,respectively.To reduce the effect of covariance on the experimental results,the screened HFI were subjected to principal component analysis(PCA)for dimensionality reduction according to the KMO test results.Finally,two methods,ridge regression and extreme learning machine(ELM),were used to construct and evaluate the PNC estimation model of each growth period of potato with VIs,HFI principal components,and VIs+HFI principal components as model variables.The results showed that:(1)HFI at different scales contributed to the estimation of PNC in each growth period of potato.(2)The accuracy and stability of the potato PNC estimation model for each growth period constructed with VIs+HFI as model variables werehigher
作者
樊意广
冯海宽
刘杨
边明博
赵钰
杨贵军
钱建国
FAN Yi-guang;FENG Hai-kuan;LIU Yang;BIAN Ming-bo;ZHAO Yu;YANG Gui-jun;QIAN Jian-guo(Key Laboratory of Quantitative Remote Sensing in Agriculture of Ministry of Agriculture and Rural Affairs,Information Technology Research Center,Beijing Academy of Agriculture and Forestry Sciences,Beijing 100097,China;Nanjing Agricultural University,National Engineering and Technology Center for Information Agriculture,Nanjing 210095,China;National Engineering Research Center for Information Technology in Agriculture,Beijing 100097,China;Key Laboratory of Modern Precision Agriculture System Integration Research,Ministry of Education,China Agricultural University,Beijing 100083,China;School of Mapping and Geographical Science,Liaoning Technical University,Fuxin 123000,China)
出处
《光谱学与光谱分析》
SCIE
EI
CAS
CSCD
北大核心
2023年第5期1532-1540,共9页
Spectroscopy and Spectral Analysis
基金
黑龙江省揭榜挂帅科技攻关项目(2021ZXJ05A05)
国家自然科学基金项目(41601346)
2022年度农业农村部农业遥感机理与定量遥感重点实验室建设项目(PT2022-24)资助。
关键词
无人机
马铃薯
植株氮含量
植被指数
高频信息
Unmanned aerial vehicle
Potato
Plantnitrogen content
Vegetation indices
High frequency information
