| 陈震,程千,徐洪刚,黄修桥.不同水肥处理下夏玉米株高、生物量响应特征及光谱反演[J].干旱地区农业研究,2023,(4):198~207 |
| 不同水肥处理下夏玉米株高、生物量响应特征及光谱反演 |
| Inversion model of summer maize plant height and biomass under different water and fertilizer treatments based on UAV spectra |
| |
| DOI：10.7606/j.issn.1000-7601.2023.04.21 |
| 中文关键词: 夏玉米 无人机 光谱 水肥管理 机器学习 |
| 英文关键词:summer maize UAV spectrum water and fertilizer management machine learning |
| 基金项目:“十四五”国家重点研发计划课题（2022YFD1900404）；中国农业科学院科技创新工程 |
|
| 摘要点击次数: 939 |
| 全文下载次数: 714 |
| 中文摘要: |
| 以不同水肥处理下的大田夏玉米为研究对象，利用无人机遥感系统采集光谱数据，分析光谱数据及地面测量的玉米生长指标，明确不同水肥处理对夏玉米生长的影响，构建了基于光谱感知的作物生长指标监测模型。结果表明，不同灌溉施肥处理显著影响夏玉米株高，光谱计算的株高值与实测值均在P<0.0001水平上极显著相关，2020年拔节期的2个时段和喇叭口期的2个时段决定系数R2分别为0.354、0.483、0.672、0.702，2021年拔节期、喇叭口期、抽雄期和吐丝期R2分别为0.314、0.410、0.426、0.466。多个生育时期数据融合可以大幅提高光谱反演株高的精度，两年的拟合优度分别为0.946和0.906。多光谱植被指数与不同水肥处理下的夏玉米生物量相关性较好，利用Cubist算法构建的2020年玉米生物量反演模型表现最优；多个生育时期数据融合可以显著提高模型的反演精度，3种算法构建的模型（SVR模型，Cubist模型和RF模型）在2021年生育时期融合数据集上均表现较优，其在测试集上的R2分别达到了0.942、0.941、0.934。 |
| 英文摘要: |
| Summer maize under different water and fertilizer treatments was used as the main line of research. Images were collected by using UAV remote sensing system. By analyzing the spectral data and the maize growth indexes measured on the ground, we clarify the effects of different water and fertilizer treatments on the growth indexes of summer maize and construct a crop growth index monitoring model based on spectral perception.The results showed that different irrigation and fertilization treatments significantly affected summer maize plant height, and the spectrally calculated plant height values were all highly significantly correlated with the measured values at the P<0.0001 level. R2 of the coefficient of determination for different periods in 2020 were 0.354, 0.483, 0.672, and 0.702, respectively and in 2021 were 0.314, 0.410, 0.426, and 0.466, respectively. Growth period data fusion can greatly improve inversion accuracy of plant height, with goodness of fit of 0.946 and 0.906, respectively in 2020 and 2021. There was a good correlation between multi\|spectral vegetation index and summer maize biomass under different water and fertilizer treatments. The inversion model for 2020 maize biomass constructed using the Cubist algorithm performed best. Growth period data fusion greatly improved the inversion superiority of the model. Three models （SVR, Cubist and RF model） constructed by the three algorithms all had excellent performance on the 2021 data set, with R2 on the test set reaching 0.942, 0.941 and 0.934, respectively. |
| 查看全文 查看/发表评论 下载PDF阅读器 |