基于无人机多光谱遥感的棉花脱叶效果监测及处方图生成研究  

Study on cotton defoliation effect monitoring and formulation map generation based on multi-spectral remote sensing of UAV

作　　者：闫春雨 黎文华 兰玉彬 胡传旭 鲁力群[2,3] 赵静[1,2] YAN Chunyu;LI Wenhua;LAN Yubin;HU Chuanxu;LU Liqun;ZHAO Jing(School of Agricultural Engineering and Food Science,Shandong University of Technology,Zibo 255049,China;International Precision Agriculture Aviation Application Technology Research Center,Shandong University of Technology,Zibo 255049,China;School of Transportation and Vehicle Engineering,Shandong University of Technology,Zibo 255049,China)

机构地区：[1]山东理工大学农业工程与食品科学学院,山东淄博255049 [2]山东理工大学国际精准农业航空应用技术研究中心,山东淄博255049 [3]山东理工大学交通与车辆工程学院,山东淄博255049

出　　处：《南京农业大学学报》2022年第4期799-808,共10页Journal of Nanjing Agricultural University

基　　金：山东省引进顶尖人才“一事一议”专项(鲁政办字〔2018〕27号);国家现代农业产业技术体系项目(CARS-15-22)。

摘　　要：[目的]本研究旨在解决人工监测棉花脱叶催熟效果耗时、费力等问题。[方法]利用四旋翼无人机获取喷施棉花脱叶剂前后4次多光谱图像,采用Pix4Dmapper软件拼接无人机图像,计算土壤调整植被指数(SAVI)、比值植被指数(RVI)、差值植被指数(DVI)和归一化植被指数(NDVI)4种植被指数,利用最大熵阈值法和植被指数阈值法提取棉叶覆盖信息。利用支持向量机对喷施脱叶剂的棉花进行监督分类,对4次采集的多光谱图像总体分类精度均大于97%,Kappa系数均大于0.95,因此将支持向量机分类结果作为真值,对最大熵阈值法和基于植被指数阈值法提取的棉花脱叶信息进行验证。将最佳提取方法用于建立棉花脱叶效果监测模型,代替人工监测棉花脱叶效果。根据最佳脱叶效果监测模型制作第2次脱叶剂施药处方图,指导第2次脱叶剂的变量喷施。[结果]在整个棉花脱叶过程中,基于SAVI_(840)植被指数阈值法监测棉花脱叶效果优于基于RVI_(940)最大熵阈值法,前者为最优监测棉花脱叶效果的模型,将最优监测模型提取的结果与田间调查棉叶数拟合,对数模型的R^(2)最高,为0.96,说明无人机遥感监测棉花脱叶效果可行;根据最优监测模型提取棉叶信息制作变量喷施处方图进行施药并验证,结果表明施药效果较好,与常规定量施药相比,可节约农药7.39%,最高节药率达14.61%。[结论]无人机遥感技术可以代替人工大面积、快速和准确监测棉花脱叶效果,利用监测结果生成的处方图进行变量施药,可实现减药增效。[Objectives]The target of research was to solve time-consuming and laborious problems when monitoring the effect of cotton defoliation and ripening with labor.[Methods]A quad rotor UAV(unmanned aerial vehicle)was used to obtain four multispectral images before and after spraying cotton defoliant.The UAV images were spliced by Pix4Dmapper software.The four planting indexes were calculated including the soil adjusted vegetation index(SAVI),the ratio vegetation index(RVI),difference vegetation index(DVI),and normalized difference vegetation index(NDVI).The cover information of cotton leaf was extracted by methods of the maximum entropy threshold and vegetation index threshold.The support vector machine was used to supervise the classification of cotton sprayed with defoliants.The overall classification accuracy of the multispectral images collected for four times was greater than 97%,and the Kappa coefficients were all greater than 0.95.Therefore,the results of support vector machine classification were used as the true value.The information of cotton defoliation extracted with the methods of the maximum entropy threshold and the vegetation index threshold was verified.The monitoring model of cotton defoliation effect was established with the optimal extraction method.According to the best monitoring model of defoliation effect,the prescription map of the second defoliant spraying was made to guide the variable spraying of the second defoliant.[Results]The monitoring effect of cotton defoliation based on SAVI_(840) vegetation index threshold method was better than that based on RVI_(940) maximum entropy threshold method in the whole process of cotton defoliation.The former was the optimal model for monitoring cotton defoliation effect.The results were fitted with the number of cotton leaves in the field survey,and the logarithmic model had the highest R 2 of 0.96,which showed the feasible effect of remote sensing monitoring of cotton defoliation by UAV.According to the optimal monitoring model,the cotton leaf inform

关 键 词：无人机遥感 棉花脱叶效果 监测模型 处方图 变量施药 

分 类 号：P237[天文地球—摄影测量与遥感]