西南农业学报2025，Vol.38Issue(10)：2108-2119,12.DOI:10.16213/j.cnki.scjas.2025.10.007
基于无人机多光谱数据的冬播春小麦产量预测
Winter-sownspring wheat yield prediction based on multispectral data from unmanned aerial vehicles
摘要
Abstract
[Objective]The study developed a vegetation index-based winter-sown spring wheat yield prediction model under different water and fertilizer treatments to achieve efficient and accurate yield estimation,providing critical technical support and theoretical foundation for precision agricultural management in Xinjiang,thereby advancing the region's agricultural modernization.[Method]Using Xin Chun 44 wheat as the experimental subject,a split-plot design was employed.The main plot parameters included irrigation water([3000 m3/hm2(W1),3600 m3/hm2(W2),4200 m3/hm2(W3)]and nitrogen fertilizer application[0 kg/hm2(N0),120 kg/hm2(N1),240 kg/hm2(N2)].A DJI MAVIC 3 Pro drone captured multispectral imagery during five critical growth stages[jointing,panicle initiation,flowering stage,10-day post-flowering(grain filling early)and maturation stage].Through image processing to extract spectral band data values and field measure-ments of yields,we analyzed correlations with vegetation indices and identified the optimal vegetation index for constructing a winter-sown spring wheat yield estimation model.[Result]Significant differences were observed in the correlation between vegetation indices and yields across individual growth stages.The correlation coefficients were the highest during the jointing stage(NDVI and OSAVI,r=0.832),panicle initiation stage(GNDVI,r=0.849),flowering stage(GNDVI,r=0.916),10-day post-flowering period(RDVI,r=0.930),and maturity stage(CIRE,r=0.868).The 10-day post-flowering training set achieved the highest R2(0.865),while the flowering validation set demon-strated the best R2(0.818).The full growth cycle model showed significantly better accuracy than single-phase models(RMSE range 270.1-578.6 kg/hm2).[Conclusion]The optimal time window for growth parameter prediction in winter-sown spring wheat is from flowering to pre-maturity stages(particularly 10 days post-flowering).Integrating multi-phase vegetation index data significantly enhances model accuracy and stability.关键词
春小麦/冬播/无人机/植被指数/产量Key words
Spring wheat/Winter sowing/Drone/Vegetation index/Yield分类
农业科技引用本文复制引用
蒋鹏程,薛丽华,冯豪博,田泽,章建新,赵小宁..基于无人机多光谱数据的冬播春小麦产量预测[J].西南农业学报,2025,38(10):2108-2119,12.基金项目
新疆农业科学院自主培育专项项目(xjnkyzzp-2022001) （xjnkyzzp-2022001）
国家自然科学基金项目(32060433) （32060433）
