| 连续投影与SSA-ELM结合的玉米氮平衡指数高光谱估测 |
| 点此下载全文 |
| 引用本文：郭 松，常庆瑞 ，张佑铭，陈 倩，落莉莉.连续投影与SSA-ELM结合的玉米氮平衡指数高光谱估测[J].西北农业学报,2023,(1):130~138 |
| DOI:10.7606/j.issn.1004-1389.2023.01.014 |
| 摘要点击次数: 123 |
| 全文下载次数: 39 |
|
| 基金项目:国家高技术研究发展计划（863计划）（2013AA102401）。 |
|
| 中文摘要:玉米氮平衡指数的快速、无损估测，对监测植株长势和指导田间氮肥施用具有重要现实意义。通过田间试验观测各生育期（拔节期、抽雄期、乳熟期、完熟期）玉米叶片氮平衡指数和对应光谱反射率，应用特征波段和植被指数，构建不同类型的玉米氮平衡指数高光谱反演模型。结果表明：不同氮平衡指数下的叶片高光谱反射特征基本一致，在可见光波段反射率较低、近红外波段反射率较高，不同生育期氮平衡指数与光谱反射率的相关性在抽雄期最高，乳熟期、完熟期次之、拔节期最低；连续投影算法具有良好的降维效果，筛选出的各生育期建模光谱参数为8~20个；各生育期多因素模型均优于单因素模型，抽雄期各类模型效果最佳，其中基于麻雀搜索算法优化的极限学习机回归模型是最优模型，其建模R与验证R均为0.93，相对预测偏差分别为2.57、2.31，表明该模型对氮平衡指数具有极好的预测能力。 |
| 中文关键词:玉米 氮平衡指数 连续投影算法 极限学习机回归 |
| |
| Hyperspectral Estimation of Maize Nitrogen Balance Index by Successive Projection Combinedwith SSA-ELM |
|
|
| Abstract:Fast and nondestructive estimation for maize nitrogen balance index (NBI) is of great importance to monitor growth state and guide field nitrogen fertilization. We firstly observed leaf NBI and corresponding spectral reflectance of maize at four growth stages (jointing stage, tasseling stage, milk-ripe stage, and full ripeness stage) in the fields, then, constructeda group of maize NBI hyperspectral inversion models based on the selected original spectral features or vegetation indices, and their performance were consequently compared to obtain the optimal NBI inversion model at different growth stages. The experiment results showed that the leaves with different NBI basically shared the similar hyperspectral reflectance patterns,reflectance in visible bands were lower than that in near infrared band, and the highest correlation between the NBI and spectral reflectance appeared at tasseling stage, followed by milk-ripe and full-ripe stages, and the lowest one appearedat jointing stage. The successive projection algorithm performed well on dimensionality reduction, and the number of the optimal spectral features for each growth stage rangeed from 8 to 20. Multi-factor models outperform single factor models at all growth stages, and the models at tasseling stage were prior to that at other three growth stages. Among the inversion models at tasseling stage, extreme learning regression model optimized with sparrow search algorithm had the best performance, both the modeling R and the validation R were 0.93, and the relative prediction deviation were 2.57, 2.31, respectively. This studydemonstrated that extreme learning regression model optimized with sparrow search algorithm has an excellent forecasting in maize NBI. |
| keywords:Maize Nitrogen balance index Successive projections algorithm Extreme learning machine regression |
| 查看全文 查看/发表评论 下载PDF阅读器 |
