北方农业学报 ›› 2024, Vol. 52 ›› Issue (4): 120-134.doi: 10.12190/j.issn.2096-1197.2024.04.13

• 农业信息技术 • 上一篇    

基于无人机多光谱影像的春小麦SPAD值估测研究

张鹏, 王砚涵, 吴强, 李赛如, 张永平   

  1. 内蒙古农业大学 农学院, 内蒙古 呼和浩特 010019
  • 收稿日期:2024-04-24 出版日期:2024-08-20 发布日期:2024-11-22
  • 通讯作者: 张永平(1970—),男,教授,博士,博士生导师,主要从事寒旱区作物栽培生理生态、作物高产优质高效栽培、小麦水肥高效利用与精确栽培的研究工作。
  • 作者简介:张 鹏(1998—),男,硕士研究生,主要研究方向为作物栽培。
  • 基金资助:
    内蒙古自治区直属高校基本科研业务费项目(BR231509);内蒙古自治区高等学校碳峰值碳中和研究专项项目(STZX202214);内蒙古自治区小麦产业技术创新推广体系

Research on SPAD value estimation of spring wheat based on UAV multispectral imager

ZHANG Peng, WANG Yanhan, WU Qiang, LI Sairu, ZHANG Yongping   

  1. College of Agriculture, Inner Mongolia Agricultural University, Hohhot 010019, China
  • Received:2024-04-24 Online:2024-08-20 Published:2024-11-22

摘要: 【目的】探究无人机遥感技术对春小麦SPAD值的无损估测,建立小田块尺度上的春小麦SPAD值估测模型。【方法】利用多光谱无人机获取2022年内蒙古河套地区春小麦冠层多光谱数据,并提取5种植被指数,通过分析植被指数与植株SPAD值的相关性,结合支持向量机回归(SVR)、偏最小二乘回归(PLSR)和随机森林回归(RFR)构建春小麦关键生育时期SPAD值估测模型,并筛选各时期SPAD值最优估测模型。【结果】通过与5种植被指数配对来确定最佳模型,在春小麦的拔节期、抽穗期、开花期和灌浆期,这15种组合方法均可对SPAD值进行估测。对3种建模方法的最优模型进行比较发现,NDVI+RFR模型在估测春小麦SPAD值方面具有最高的准确性和良好的鲁棒性,其R2=0.96,RMSE=0.19。【结论】利用无人机估测春小麦的SPAD值是可行的,最优模型为NDVI+RFR模型。

关键词: 无人机监测, 多光谱影像, 春小麦, SPAD值

Abstract: 【Objective】To study the non-destructive estimation of spring wheat SPAD value by UAV remote sensing technology,and establish a model for spring wheat SPAD value estimation on a small field scale.【Methods】A multispectral UAV was used to acquire spring wheat canopy multispectral data in 2022 in Hetao region of Inner Mongolia,and 5 vegetation indices were extracted. The correlation between vegetation indices and plant SPAD values was analyzed,and support vector regression(SVR),partial least squares regression(PLSR),and random forest regression(RFR) were combined to construct SPAD value estimation models for key growth periods of spring wheat. The optimal SPAD value estimation models for each period were selected.【Results】The optimal model was determined by pairing with 5 vegetation indices,and all 15 combination methods were able to estimate SPAD values during the jointing,heading,flowering,and filling stages of spring wheat. Comparing the optimal models of 3 modeling methods,it was found that the NDVI+RFR model had the highest accuracy and good robustness in estimating the SPAD value of spring wheat,with R2=0.96 and RMSE=0.19.【Conclusion】It is feasible to use UAV to estimate the SPAD value of spring wheat, and the optimal model is the NDVI+RFR model.

Key words: UAV monitoring, Multispectral imaging, Spring wheat, SPAD value

中图分类号: 

  • S512.12