无人机成像高光谱‘时-空-谱’融合耦合作物模型的冬小麦氮素营养状况预测

Early diagnosis of crop nitrogen deficiency is very important for the timely establishment of the correct fertilization program, reducing the damage and loss of nitrogen stress on crops. The project focuses on crop nitrogen nutrition status prediction. There are three main problems in crop nitrogen nutrition prediction using remote sensing technology: (1) Multi-spectral remote sensing data lack crop nitrogen stress diagnostic features;(2) Existing hyperspectral remote sensing data from satellite-borne and manned platforms cannot simultaneously take into account high spatial-temporal resolution to meet crop nitrogen nutrition monitoring;(3) It lacks prediction model for nitrogen nutrition status in key crop growth stages. In this proposal, the nitrogen nutrient index was used as an evaluation index of nitrogen nutrition status. With field experiments of winter wheat, the main work of this study are as follows: (1) Developing “spatial-spectral”fusion algorithm for UAV (Unmanned Aerial Vehicle) based snap-shot hyperspectral images; (2) Mining spatial, spectral and temporal features of hyperspectral images and studying feature fusion method to improve the estimation accuracy of crop parameter models, and to reduce the uncertainty problem of parameter inversion only relying on spectral information; (3) Developing assimilation mechanisms and couple method for instantaneous remote sensing information and crop growth model and establish continuous spatio-temporal crop nitrogen nutrition status inversion model to achieve continuous dynamic monitoring and prediction of crop nitrogen nutrition status. The implementation of this project will help to expand the crop nitrogen nutrition diagnosis from the "instantaneous remote sensing" to "continuous phase". It is of great significance for the diagnosis of early nitrogen deficits in crops and the implementation of precision agriculture fertilization management.

提前预测作物氮素营养状况对于及时制定正确的施肥方案，减少氮胁迫对作物造成的危害和损失非常重要。本项目围绕作物氮素营养精准预测，针对目前多光谱缺少作物氮素胁迫诊断特征光谱、现有星载和有人机载平台高光谱遥感信息无法同时兼顾高时空分辨率以满足作物氮素营养监测和缺乏作物关键生育期氮素营养状况提前预测模型等问题，以冬小麦为研究对象，将氮营养指数作为氮素营养状况评价指标，发展无人机画幅式成像高光谱图谱融合算法;融合高光谱影像光谱特征与时空特征来提高作物参数模型估算精度，减少仅依赖光谱信息存在的参数反演不确定性问题;在此基础上研究遥感瞬时信息与作物生长模型结合的时空耦合和同化机理，建立时空连续的作物氮素营养信息反演模型，实现作物氮素营养的连续动态监测与预测。本项目的实施有助于实现作物氮素营养诊断从“遥感瞬时”向“时相连续”的拓展，对于作物氮素早期亏缺诊断，实施精准农业施肥管理具有重要意义。

在作物关键施肥期之前准确诊断氮素营养亏缺状况，对及时进行田间管理调控，实现精准施肥具有重要意义。本项目围绕作物氮素营养精准预测，以冬小麦为研究对象，深度挖掘了画幅式无人机成像高光谱遥感在作物氮素营养状况监测和诊断中的优势，并将遥感瞬时信息与DSSAT作物模型耦合，建立时空连续的作物氮素营养信息反演模型。主要研究内容包括:(1)分析并评估了三类“空-谱”融合算法，即成分替换、多分辨率分析和基于子空间的方法，确定了适合作物氮素监测的高光谱影像融合方法;(2)针对如何有效选取合理指标来评判作物氮素营养水平的问题，从冠层高光谱响应的角度，采用全局敏感性分析和植被指数相关性方法，给出了不同生育阶段氮素指标的选取规则;(3)构建了一个新的用于估算作物氮营养指数(Nitrogen Nutrition Index, NNI)的遥感指数NNIRS=CIred edge/(a×sLAIDI^b)，该指数稳定性和估算性能均较好。从光谱空间域和频率域挖掘了与作物氮素相关的光谱特征，并融合空间纹理特征，构建了作物氮素和生物量估算模型，减少了仅依赖光谱信息存在的参数反演不确定问题;(4)构建了DSSAT与遥感作物生物量和植株冠层氮累积的双变量同化策略，实现了作物氮素营养状况的连续动态监测与预测。基于以上研究结果，已发表学术论文7篇，其中SCI论文6篇(包括中科院1区TOP论文3篇，2区论文2篇)。项目组晋升中级职称1人，项目培养博士后2名。综上，本项目按照项目研究计划，完成了各项预期目标。