|
| 基于无人机高光谱影像的田块尺度玉米估产与生育时期优选 |
| Field-scale maize yield estimation and growth stages optimization based on UAV hyperspectral image |
| Received:October 10, 2023 |
| DOI：10.13254/j.jare.2023.0634 |
| 中文关键词: 无人机，高光谱影像，田块尺度，玉米，产量，机器学习 |
| 英文关键词: UAV, hyperspectral image, field scale, maize, yield, machine learning |
| 基金项目:国家自然科学基金项目（42301074）；国家重点研发计划项目（2021YFD1500800） |
| Author Name | Affiliation | E-mail | | JIA Zenghui | College of Resources and Environment, Jilin Agricultural University, Changchun 130118, China
Key Laboratory of Soil Resource Sustainable Utilization of Jilin Province in Commodity Grain Bases, Changchun 130118, China | | | ZHANG Jizhen | Songliao Water Resources Commission Songliao Basin Soil and Water Conservation Monitoring Center Station, Changchun 114046, China | | | HAO Hang | College of Resources and Environment, Jilin Agricultural University, Changchun 130118, China
Key Laboratory of Soil Resource Sustainable Utilization of Jilin Province in Commodity Grain Bases, Changchun 130118, China | | | ZHANG Xingyu | College of Resources and Environment, Jilin Agricultural University, Changchun 130118, China
Jilin Emergency Warning Information Dissemination Center, Changchun 130062, China | | | XIA Chenzhen | College of Resources and Environment, Jilin Agricultural University, Changchun 130118, China
Key Laboratory of Soil Resource Sustainable Utilization of Jilin Province in Commodity Grain Bases, Changchun 130118, China | | | GAO Qiang | College of Resources and Environment, Jilin Agricultural University, Changchun 130118, China
Key Laboratory of Soil Resource Sustainable Utilization of Jilin Province in Commodity Grain Bases, Changchun 130118, China
Key Laboratory of Straw Comprehensive Utilization and Black Land Protection, Ministry of Education, Changchun 130118, China | | | ZHANG Yue | College of Resources and Environment, Jilin Agricultural University, Changchun 130118, China
Key Laboratory of Soil Resource Sustainable Utilization of Jilin Province in Commodity Grain Bases, Changchun 130118, China
Key Laboratory of Straw Comprehensive Utilization and Black Land Protection, Ministry of Education, Changchun 130118, China | lisa_ling7892002@163.com |
|
| Hits: 728 |
| Download times: 491 |
| 中文摘要: |
| 为实现东北黑土区田块尺度上玉米产量的精准估算与生育时期优选，本研究以我国东北黑土区的春玉米为研究对象，选取吉林省梨树县的长期定位玉米试验田，于2019、2020年利用无人机采集玉米3个关键生育时期（拔节期、吐丝期、成熟期）的冠层高光谱影像，选取10种与产量显著相关的窄波段植被指数，并结合作物农学参数与施肥信息，分别采用逐步回归、随机森林（RF）和极度梯度提升树（XGBoost）算法构建玉米产量估算模型。最后通过决定系数（R2）、均方根误差（RMSE）和归一化均方根误差（NRMSE）对产量模型进行精度评价，以筛选出最优估产模型。结果表明：3种产量预测模型中XGBoost模型估算精度较优，其2019年吐丝期的R2、RMSE和NRMSE分别为0.93、1 054.17 kg·hm-2和11.68%。同时，3种模型均表现为在吐丝期估算精度最优，最佳模型——2019年吐丝期的XGBoost模型中用于玉米产量估算的指示因子——植被指数R-M、作物农学参数与施肥信息的特征重要性分别为19.72%、4.70%、62.41%。研究表明，结合无人机影像与机器学习算法并融合多源辅助信息可提高田块尺度玉米产量的估算精度，为农业生产中的作物产量精准预估提供数据支撑与科学参考。 |
| 英文摘要: |
| To achieve accurate maize yield estimation and optimize growth period, the experiment was conducted at the field scale in the northeast black soil region. The study region comprised the long-term maize experimental fields situated in Lishu County, Jilin Province. Canopy hyperspectral data obtained by UAV at three key growth stages of maize（jointing, silking, maturity）were utilized to select 10 narrowband vegetation indices from 2019 to 2020. The yield estimation models at different stages were established by stepwise regression, random forest（RF）, and Extreme Gradient Boosting（XGBoost）models, combined with vegetation indices, crop characteristics, and fertilization information. Finally, the coefficient of determination（R2）, root mean square error（RMSE）, and normalized root mean square error（NRMSE）were used to evaluate the yield estimation model. The results showed that the yield estimation model based on the XGBoost method was relatively better than the models based on stepwise regression and RF. The XGBoost model showed the best performance at the silking stage in 2019（R2 = 0.93, RMSE = 1 054.17 kg·hm-2, and NRMSE=11.68%）. The contributions of the XGBoost model variables of maize yield estimation at the silking stage were 19.72%, 4.70%, and 62.41% for vegetation index R-M, crop characteristics, and fertilization information, respectively. Hence, the integration of UAV hyperspectral imaging with machine learning algorithms, alongside auxiliary information, enhances the accuracy of maize yield estimation at the field scale. This approach offers valuable support data and scientific insights for precise crop yield prediction in agricultural production. |
| HTML
View Full Text
View/Add Comment Download reader |
| Close |
|
|
|
