基于无人机可见光遥感的马铃薯植株氮素累积估测与验证

Estimation and Validation of Nitrogen Accumulation of Potato Based on UAV Visible Light Remote Sensing

为探究无人机可见光遥感技术快速、无损监测马铃薯氮素营养的可能性,于2020—2022年在贵州省威宁县开展马铃薯不同氮肥梯度大田试验,设置7个氮素水平(0、60、120、180、240、300和360 kg·hm^(-2)),利用无人机搭载可见光传感器获取不同年份马铃薯块茎形成期冠层RGB高清影像,并同步测定马铃薯地上部氮素含量、生物量和氮素累积量等氮素营养指标,以2020—2021年数据作为建模数据,以2022年数据作为验证数据,建立氮素营养指标估测方程模型并绘制实测值和预测值的1∶1线性关系图。结果表明,与其他冠层光谱参数相比,红光和蓝光比值(R/B)能更好地表征马铃薯氮素营养指标,其与地上部氮素含量、生物量、氮素累积量的相关性均达到极显著水平(P<0.01),其中二次函数模型相关性均优于其他函数模型。利用2022年相同独立试验验证该模型准确性,地上部氮素含量、生物量、氮素累积量实测值与预测值的决定系数(R2)分别为0.949、0.977、0.977,均方根误差(RMSE)和相对误差(MRE)分别为0.368、0.149、0.073和10.42%、6.05%、8.85%,表明模型预测精度较好。综上,无人机可见光遥感可用于马铃薯氮素营养的评估预测,块茎形成期最佳预测参数为R/B,二次函数模型为最佳预测模型。本研究为马铃薯氮素营养无损评估预测提供了理论与实践依据。

To explore the possibility of using unmanned aerial vehicle(UAV)visible light remote sensing technology to quickly and non-destructively monitor potato nitrogen nutrition,field experiments with different nitrogen fertilizer gradients(0,60,120,180,240,300 and 360 kg·hm^(-2))were conducted in Weining County,Guizhou Province from 2020 to 2022.Using UAV equipped with visible light sensors to obtain canopy high-definition RGB images of potato tuber formation stages in different years,and synchronously collecting potato above-ground nitrogen content,biomass and nitrogen accumulation as nitrogen nutrition indicators.Using data from 2020 to 2021 as modeling data and 2022 as validation data,an equation model was established to estimate nitrogen nutrition indicators and a 1∶1 linear relationship graph was drawn between the measured and predicted values.The results showed that compared with other canopy spectral parameters,the ratio of red and blue light(R/B)was better to characterize potato nitrogen nutrition indicators,and the correlation with above-ground nitrogen content,biomass and nitrogen accumulation had reached a highly significant level(P<0.01).Among all models,the quadratic function model exhibited better correlation than other function models.The accuracy of the model was verified using the same independent experiment in 2022.The R2 of the measured and predicted values of above-ground nitrogen content,above-ground biomass and above-ground nitrogen accumulation were 0.949,0.977 and 0.977,and the root mean square error(RMSE)and relative error(MRE)were(0.368,10.42%),(0.149,6.05%)and(0.073,8.85%),respectively,indicating that the model had good predictive accuracy.In summary,UAV visible light remote sensing could be used for the evaluation and prediction of potato nitrogen nutrition.The best prediction parameter for the tuber formation period is R/B,and the quadratic function model is the best prediction model.This study provides theoretical and practical basis for non-destructive assessment and prediction of potatoe nitrogen nutrition.