无人机多角度成像方式的饲料油菜生物量估算研究

Estimation of feed rapeseed biomass based on multi-angle oblique imaging technique of unmanned aerial vehicle

旨在探索并评估一种通过无人机平台搭载可见光相机提取饲料油菜生物量的新方法。试验于2018年在华中农业大学油菜试验基地展开,利用无人机搭载五相机倾斜摄影系统同时从多个角度获取油菜终花期的可见光图像,试验共设置3种无人机飞行高度(40、60和80 m)和3种播种密度(3.00×10^(5)、5.25×10^(5)和7.50×10^(5)株hm^(-2)),并评估和对比了多角度和单相机垂直2种成像方式的生物量预测结果。试验首先通过无人机图像提取油菜冠层覆盖度和株高信息;然后通过株高在覆盖面积上进行累加获得作物体积模型;最后基于作物体积模型与实测生物量建立线性回归模型预测油菜干物质重量。结果表明,(1)在本试验设置的3个飞行高度中,随着无人机飞行高度下降,生物量预测精度呈上升趋势,其中飞行高度为40 m时,油菜生物量估算精度最佳(校正集:r=0.792,RMSE=125.0 g m^(-2),RE=13.2%;验证集:r=0.752,RMSE=139.1 g m^(-2),RE=15.3%)。(2)种植密度越高,其实际生物量越小,通过作物体积模型预测生物量的效果更好。(3)多角度成像方式与单相机垂直成像方式在油菜生物量估测精度上没有显著差异,两者皆在40 m高度下具有最好的生物量预测效果,相关系数r分别为0.772和0.742。以上结果表明,基于无人机低成本可见光成像建模技术提取饲料油菜生物量是可行的,本研究可为大田作物地上生物量信息的无损高效监测提供易于实施的解决方案和技术参考。

To obtain above-ground biomass information quickly and accurately facilitating crop growth monitoring and yield prediction,this study was to evaluate a new method to extract the biomass of feed rapeseed based on UAV with visible-light cameras.The experiment was conducted at the rapeseed experimental base of Huazhong Agricultural University in 2018.To estimate above-ground biomass of rapeseed,a UAV(unmanned aerial vehicle)platform equipped with a five-camera oblique photography system was used to simultaneously obtain images of rapeseed during final flowering period from multiple angles.Three flight altitudes(40,60,and 80 m)and three seeding densities(3.00×10^(5),5.25×10^(5),and 7.50×10^(5) plant hm^(-2))were carried out to assess biomass predictions in a single-camera vertical imaging pattern.Firstly,the rapeseed canopy coverage and plant height information from the image of the UAV were extracted.Secondly,the volume model of rapeseed was obtained by the addition of plant height on the covering area.Finally,a linear regression model was established based on volume model and measured biomass topredict the dry weight of rapeseed.The results were as follows:(1)With the decrease of the flight height of the UAV of the three flight altitudes,the accuracy of biomass prediction was on the rise,and when the flight height was 40 meters,the accuracy of rapeseed biomass estimation was the best(calibration set:r=0.792,RMSE=125.0 g m^(-2),RE=13.2%;validation set:r=0.752,RMSE=139.1 g m^(-2),RE=15.3%).(2)When the planting density of rapeseed was higher,the actual biomass was smaller,and the prediction of biomass had a better result by volume model.(3)There was no significant difference in the accuracy of rapeseed biomass estimation between multi-angle imaging and single-camera vertical imaging,both of which had the best results at the flight height of 40 meters with correlation coefficients r of 0.772 and 0.742,respectively.This study indicated that it was feasible to obtain images for extracting rapeseed biomass by a UAV,which could provide the reference for efficient and accurate phenotypic information of field crops.