基于无人机载高光谱空间尺度优化的大豆育种产量估算

发布时间：2018-05-06 20:59

本文选题：无人机 + 遥感　；参考：《农业工程学报》2017年01期

【摘要】：为探讨无人机载高光谱空间尺度对大豆产量预测精度的影响,该文以山东嘉祥圣丰大豆为研究对象,设计以多旋翼无人机为平台搭载Cubert UHD185成像高光谱传感器的无人机遥感农情监测系统,获取了大豆多个生育期的无人机高光谱数据。首先,该研究利用盛荚期-始粒期(R4-R5期)的高光谱影像,由21个不同光谱空间尺度提取的高光谱数据构建植被指数,通过植被指数方差分析结果可知所选冠层植被指数与不同品种大豆植株的生长状况密切相关,但是不同空间尺度下的F值仍存在较为明显的差异;其次,采用偏最小二乘回归建立产量与不同空间尺度的植被指数之间的回归模型,通过模型方程估算精度的曲线变化趋势进一步将最优空间尺度面积确认至9.03~10.13 m2,即当采样空间尺度区域长、宽与小区总长、宽比例介于4.25:5和4.5:5时,所得到的冠层光谱能够尽可能准确地估测大豆产量,此时估算产量和实测产量呈极显著相关(相关系数r=0.811 7,参与建模的样本个数270)。该研究可为使用高、低空高光谱影像进行作物表型信息解析和估产提供参考。
[Abstract]:In order to investigate the influence of the spatial scale of UAV on the prediction accuracy of soybean yield, this paper designed the UAV remote sensing monitoring system with Cubert UHD185 imaging hyperspectral sensor on the platform of multi rotor unmanned aerial vehicle (UAV) in Shandong, and obtained the high spectrum number of UAV in multiple soybean growth periods. First, the study uses hyperspectral images of the podding stage (R4-R5 phase) to construct vegetation index from hyperspectral data extracted from 21 different spectral spatial scales. Through the analysis of the vegetation index variance, it is found that the selected canopy vegetation index is closely related to the growth of different varieties of soybean plants, but at different spatial scales. There are still obvious differences in the F value. Secondly, the regression model between the yield and the vegetation index of different spatial scales is established by the partial least squares regression. The optimal spatial scale area is further confirmed to 9.03~10.13 M2 by the curve variation trend of the model equation, which is when the spatial scale region is long, wide and small. When the total length, with a wide ratio of 4.25:5 and 4.5:5, the obtained canopy spectrum can estimate the yield of soybean as accurately as possible. At this time, the estimated yield and the measured yield are extremely significant (the correlation coefficient r=0.811 7, the number of samples involved in the modeling is 270). This study can be used to analyze and estimate crop phenotypic information for using high, low altitude hyperspectral images. Provide reference.

【作者单位】：北京农业信息技术研究中心;国家农业信息化工程技术研究中心;南京农业大学大豆研究所/国家大豆改良中心;
【基金】：国家自然科学基金项目(61661136003,41471285) 国家重点研发计划(2016YFD0300602) 北京市农林科学院科技创新能力建设项目(KJCX20170423)
【分类号】：S565.1;S127

【相似文献】