基于无人机多光谱遥感的黄花刺茄识别最佳物候窗口分析OA

Accurate selecting phenological stages is crucial to overcoming the bottleneck in monitoring and identifying the invasive plant Solanum rostratum against complex grassland backgrounds.To elucidate the spec-tral responses dynamics of S.rostratum during invasion and to determine its optimal monitoring phenological window,this study focused on S.rostratum in the typical steppe of Bairin Right Banner,Inner Mongolia Autonomous Region,and conducted multi-temporal monitoring at five key phenological stages:seedling,early flowering,full flowering,green fruit,and senescence.Using anmanned aerial vehicle(UAV)-based multispec-tral remote sensing imagery,the identification accuracies of S.rostratum at different phenological stages were compared.The results revealed that:(1)The identification accuracy exhibited a unimodal pattern along the phe-nological progression,with the best performance achieved at the early flowering stage,reaching an overall accu-racy of 92.2%;(2)Significantly differences in spectral separability existed among phenological stages,and canopy reflectance changes triggered by inflorescence emergence were one of the primary drivers of accuracy variation;and(3)Feature contribution analysis showed that the Near-Infrared Reflectance of Vegetation(NIRv)and Kernel Normalized Difference Vegetation Index(kNDVI)played critical roles in distinguishing S.rostratum from background features,effectively mitigating the interference of mixed pixels in complex grassland scenes.In conclusion,the rational selection of phenological stages is a key factor for UAV-based monitoring of the inva-sive plants S.rostratum,and these findings can provide a scientific basis for temporal phase selection in large-scale monitoring and early warning of invasive plants.