基于无人机遥感图像的苎麻产量估测研究

本研究旨在探索一种利用无人机-RGB系统提取的苎麻株高和可见光图像光谱信息估测产量的新方法。试验于2019年在湖南农业大学耘园苎麻基地进行,利用无人机搭载高清数码相机获取二季苎麻苗期和成熟期的图像。首先利用Pix4Dmapper生成苎麻冠层2个生育期的数字表面模型和高清数码正射图像;然后基于数字表面模型采用"差分法"计算试验小区的平均株高(DSM-based H);基于正射图像提取试验小区RGB通道均值,进而计算遥感图像数码变量和植被指数,分析苎麻种质间的图像光谱表型性状和产量株高比性状的差异性与多样性;最后采用逐步回归方法建立苎麻产量预测模型,并对各项产量解释因子进行相关性分析。结果表明:(1)基于无人机-RGB系统遥感株高与实测株高显著相关(r=0.90),修正遥感株高的均方根误差为0.04 m。(2)苎麻产量与株高信息存在极显著相关性(r=0.91),而与图像光谱表型相关性不明显。(3)融合遥感图像株高和种质特征差异构建的苎麻产量估测模型精度较高, R²=0.85, RMSE=0.71。因此,基于无人机遥感图像的苎麻产量估测是可行的,这对苎麻种质特征评价和产...

Estimation of ramie yield based on UAV (Unmanned Aerial Vehicle) remote sensing images

This paper provides a new method to estimate ramie yield by integrating plant height and germplasm characteristics obtained by UAV-RGB system. The experiment was carried out in the ramie experimental area of Yunyuan base of Hunan Agricultural University in 2019, and the images of ramie in the seedling and mature stages were obtained by using a high-definition digital camera mounted on a drone. Firstly, Pix4D mapper was used to generate the digital surface model and ortho-image of ramie canopy in two growth periods. Based on the DSM, we used “difference method” to calculate the average plant height of the experimental plot. RGB channel mean value of experimental plot was extracted based on orthography, and then digital image variables and vegetation index were calculated. Then, the difference and diversity of spectral phenotypic characters and yield/plant height ratio characters among the germplasm of ramie were analyzed. Finally, stepwise regression method was used to establish the ramie yield prediction model, and correlation analysis was carried out for each yield interpretation factor. There was a significant correlation between DSM-based H and the measured plant height (r = 0.90), with RMSE of 0.04 for the linear model established based on the corrected plant height and the measured plant height. Plant height information was significantly correlated with yield (r = 0.91), while spectral phenotype information was not significantly correlated with yield. The ramie yield prediction model established by the fusion of plant height and germplasm characteristics was highly accurate, with R² of 0.85 and RMSE of 0.71. Therefore, this study has important practical significance for resource management and yield estimation of ramie germplasm.