摘要
快速准确地获取作物的植株氮含量(PNC)信息,是农业精细化管理的关键和数字农业发展的研究热点。近年来,随着无人机和传感器技术的发展,利用多种传感器信息监测作物理化参数逐渐引起国内外学者的关注。以马铃薯为研究对象,首先,基于无人机获取了马铃薯现蕾期、块茎形成期、块茎增长期、淀粉积累期和成熟期的高光谱影像和数码影像,同时采集各生育期的地面数码影像,并实测了株高(H)、PNC和11个地面控制点(GCPs)的三维空间坐标。其次,利用无人机数码影像结合GCPs生成试验区域的数字表面模型(DSM),分别从无人机数码影像和DSM中提取马铃薯的地面覆盖度(VC)和株高(H),并利用地面数码影像计算的覆盖度(VC)和实测H验证提取的VC和H的精度。然后,根据高光谱反射率数据计算绿边参数(GEPs),构造GEPs×H×VC,GEPs/(1+VC),(GEPs+VC)×H和GEPs/(1+H)4种融合特征参数(FFPs),对高光谱影像信息和数码影像信息进行融合。最后,将各生育期提取的GEPs和构造的FFPs分别与PNC作相关性分析,筛选最优绿边参数(OGEP)和最优融合特征参数(OFFP)构建5个生育期的PNC线性估算模型,并根据相关性较高的GEPs和FFPs利用偏最小二乘(PLSR)和人工神经网络(ANN)2种回归方法构建PNC的多参数估算模型,结果表明:(1)基于无人机数码影像提取的H和VC具有较高的精度,可以代替实测H和VC估算作物理化参数。(2)与GEPs相比,前4个生育期,构造的大部分FFPs与PNC的相关性更高,能更好地反映马铃薯的氮营养状况。(3)马铃薯5个生育期,OFFP估算PNC的效果优于OGEP。(4)与单参数模型相比,基于GEPs和FFPs利用PLSR和ANN 2种方法构建的模型精度和稳定性均明显提高,其中,以FFPs为模型因子利用ANN方法构建的模型效果最好。该研究表明融合高光谱绿边参数和高清数码相机传感器提取的株高和覆盖度信息能显著提升PNC的估算精度,可为马铃薯氮营养状况的动态无损监测和多源传感器信息的应用提供参考。
Acquiring the plant nitrogen content(PNC)information of crops quickly and accurately is the key to agricultural meticulous management and a research hotspot in the development of digital agriculture.In recent years,with the development of UAV and sensor technology,the use of various sensor information to monitor the physical and chemical parameters of crops has gradually attracted the attention of scholars at home and abroad.This study takes potato as the research object.Firstly,based on the UAV,the hyperspectral images and digital images of the potato budding stage,tuber formation stage,tuber growth stage,starch accumulation stage and maturity stage were obtained.At the same time,the digital camera was used to synchronously obtain the ground digital images of five growth periods,and the three-dimensional spatial coordinates of eleven ground control points(GCPs)and plant height(H),PNC were measured.Secondly,the digital surface model(DSM)of the test area was generated by using UAV digital images combined with GCPs.The accuracy of the extracted VCand His verified by the calculated coverage(VC)of the digital image and the measured H.Then,the green edge parameters(GEPs)were calculated according to the hyperspectral images,and four fusion feature parameters(FFPs)of GEPs×H*VC,GEPs/(1+VC),(GEPs+VC)×Hand GEPs/(1+H)were constructed,fusion of hyperspectral image information and digital image information.Finally,the correlation between GEPs extracted and FFPs constructed in each growth period with PNC were analyzed,and the PNC linear estimation models of five growth periods were constructed based on the optimal GEP and optimal FFP respectively.According to the GEPs and FFPs with high correlation,the multiple parameters estimation models of PNC were constructed by using partial least squares(PLSR)and artificial neural network(ANN).The results show that:(1)Hand VCextracted from UAV digital images have high accuracy,which can replace the measured H and VC to estimation physical and chemical parameters(2)Compared with GEPs,most of the constructed FFPs have stronger correlation with PNC in the first four growth stages,and could better reflect the nitrogen nutrition status of potato.(3)Linear estimation models of potato PNC were constructed based on the optimal green edge parameter(OGEP)and the optimal fusion feature parameter(OFFP),respectively.The results showed that the effect of OFFP in estimating PNC was better than that of OGEP.(4)Compared with the single-parameter model,the accuracy and stability of the model constructed by using PLSR and ANN based on GEPs and FFPs are significantly improved.Among them,the models constructed with FFPs as the model factor have the best effect.(5)The ANN method is better than the PLSR method in estimating PNC in each growth period.Therefore,the fusion of the hyperspectral green edge parameters and the plant height and coverage information extracted by the high-definition digital camera sensor can improve the estimation accuracy of PNC,which provide a reference for the non-destructive dynamic monitoring of potato nitrogen nutrition status and the application of multi-source sensors information.
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,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 Geomatics,Liaoning Technical University,Fuxin 123000,China)
出处
《光谱学与光谱分析》
SCIE
EI
CAS
CSCD
北大核心
2022年第10期3217-3225,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(PT202224)资助。
关键词
植株氮含量
无人机
多源传感器
绿边
株高
覆盖度
Plant nitrogen content
UAV
Multi-source sensor
Green edge
Plant height
Coverage
