基于无人机遥感不同施氮量下马铃薯生长指标的估测与验证

李广鑫; 曹忠地; 刘骑郡; 张明聪; 包东庆; 姜丽丽

李广鑫,曹忠地,刘骑郡,张明聪,包东庆,姜丽丽.基于无人机遥感不同施氮量下马铃薯生长指标的估测与验证[J].干旱地区农业研究,2026,(4):261~271

基于无人机遥感不同施氮量下马铃薯生长指标的估测与验证

Estimation and validation of potato growth indicators based on unmanned aerial vehicle remote sensing under different nitrogen application rates

DOI：10.7606/j.issn.1000-7601.2026.04.25

中文关键词: 马铃薯无人机遥感施氮量生长指标归一化植被指数预测模型

英文关键词:potato UAV remote sensing nitrogen rate growth indicators NDVI prediction models

基金项目:黑龙江省“双一流”学科协同创新成果建设项目“粮食作物绿色低碳”（LJGXCG2022-107）

作者	单位
李广鑫	黑龙江八一农垦大学农学院,黑龙江大庆 163319；农业农村部东北平原农业绿色低碳重点实验室,黑龙江大庆 163319
曹忠地	黑龙江八一农垦大学农学院,黑龙江大庆 163319；农业农村部东北平原农业绿色低碳重点实验室,黑龙江大庆 163319
刘骑郡	黑龙江八一农垦大学农学院,黑龙江大庆 163319；农业农村部东北平原农业绿色低碳重点实验室,黑龙江大庆 163319
张明聪	黑龙江八一农垦大学农学院,黑龙江大庆 163319；农业农村部东北平原农业绿色低碳重点实验室,黑龙江大庆 163319
包东庆	北大荒集团黑龙江克山农场有限公司,黑龙江克山 161600
姜丽丽	黑龙江八一农垦大学农学院,黑龙江大庆 163319；农业农村部东北平原农业绿色低碳重点实验室,黑龙江大庆 163319

摘要点击次数: 0

全文下载次数: 0

中文摘要:

为解析不同施氮量下马铃薯生长指标与归一化植被指数（NDVI）之间的动态关联，构建精准的生长指标预测模型并优化氮肥管理策略，以‘Innovator’马铃薯为供试品种，设置4个氮肥处理（100、150、200、250 kg·hm^-2），通过无人机多光谱遥感技术获取马铃薯全生育期冠层NDVI数据，同步实测株高、叶绿素含量（SPAD）及产量数据，采用一元线性回归、偏最小二乘回归（PLSR）、支持向量回归（SVR）和随机森林（RF）模型分析变量间的关联性，量化施氮量对马铃薯生长参数及产量的调控效应。结果表明，施氮量对马铃薯的株高、SPAD值和产量均具有显著调控作用，4个不同生育时期的株高均在施氮量250 kg·hm^-2时达到最高；现蕾期、块茎形成期和淀粉积累期的SPAD值均在施氮量200 kg·hm^-2时最高；产量随施氮量增加呈先升后降趋势，施氮量200 kg·hm^-2时产量最高（44.3 t·hm^-2）。预测模型精度受生育阶段影响显著，现蕾期与块茎形成期的株高提取精度较高，线性回归预测模型的决定系数R²达0.89；块茎形成期为SPAD值预测的最优建模窗口期；采用RF模型融合NDVI与株高构建的SPAD值预测模型表现最佳，其R²为0.76，均方根误差RMSE为1.02。综上，基于无人机遥感（UAV）构建的预测模型可有效量化大田马铃薯生长状况，为马铃薯产量预测及氮素管理优化提供科学依据。

英文摘要:

To elucidate the dynamic relationships between potato growth indicators and the normalized difference vegetation index (NDVI) under different nitrogen application rates, to construct accurate prediction models for growth indicators, and to optimize nitrogen fertilizer management strategies, the potato variety ‘Innovator’ was grown under four nitrogen treatments (100, 150, 200, 250 kg·hm^-2). Canopy NDVI data across the whole growth period were acquired using UAV\|based multispectral remote sensing, while plant height, chlorophyll content (SPAD), and yield were synchronously measured in the field. Unary linear regression, partial least squares regression (PLSR), support vector regression (SVR), and random forest (RF) models were employed to analyze the correlations among variables and to quantify the regulatory effects of nitrogen application rates on potato growth parameters and yield. The results showed that nitrogen application rate significantly regulated plant height, SPAD value, and yield of potato. Plant height at four different growth stages reached the maximum under 250 kg·hm^-2. SPAD values at the budding period, tuber initiation stage, and starch accumulation stage were highest under 200 kg·hm^-2 nitrogen application rate. Yield first increased and then decreased with increasing nitrogen application rate, with the maximum yield (44.3 t·hm^-2) observed at 200 kg·hm^-2. The accuracy of prediction models was significantly affected by growth stage. Higher extraction accuracy for plant height was achieved at the budding period and tuber initiation stage, with the coefficient of determination R² of the linear regression prediction model reaching 0.89. The tuber initiation stage was identified as the optimal window for SPAD value prediction. The SPAD prediction model constructed by integrating NDVI and plant height using the RF model performed best, achieving an R² of 0.76 and a root mean square error RMSE of 1.02. In conclusion, the prediction model based on UAV remote sensing can effectively quantify the growth status of field\|grown potato, providing a scientific basis for potato yield prediction and optimization of nitrogen management.

查看全文查看/发表评论下载PDF阅读器