南方丘陵区林下植被覆盖度无人机多角度遥感测量

UAV Multi Angle Remote Sensing Quantification of Understory Vegetation Coverage in the Hilly Region of South China

林下植被覆盖度作为描述森林生态系统健康的重要参量,是探查水土流失情况的重要指征。量化南方丘陵地带地形起伏区林下植被覆盖度对水土流失精准治理具有重要意义。选取长汀县河田镇作为研究区,基于无人机多角度(0°、10°、20°、30°)遥感数据,利用局部最大值法,将植被形态划分为离散型与连续型,并分别利用点云布料滤波与地图森林密度算法反演出对应植被形态的林下地形。采用半高斯拟合(HAGFVC)法分离植被,再设置高程差阈值提取出林下植被,结合不同角度影像,利用形态学中的斑块膨胀方法弥补树冠遮挡的误差,进而量化林下植被覆盖度。结果表明,针对不同植被形态,上述技术路线所得的地形反演结果精度较高,其中离散型样地RMSE为0.400 m,连续型样地RMSE为0.518 m。单一角度中,基于20°影像估算林下植被覆盖度的结果精度最高(R^(2)=0.459,RMSE=0.166);基于30°影像的估算精度最低(R^(2)=0.337,RMSE=0.243)。多角度影像耦合(0°+10°+20°+30°)后,林下植被覆盖度的估算精度得到明显提高,R^(2)达到0.675,RMSE为0.102。研究结果可以为南方红壤区林下植被信息无人机遥感调查提供技术支撑。

As an important parameter to describe the forest ecosystem health,understory vegetation coverage is an important indicator to explore the situation of soil and water loss.Quantifying the understory vegetation coverage in the hilly area of south China is of great significance to the accurate control of soil and water loss.In this paper,Hetian Town,Changting County was selected as the study area.Based on the UAV multi angle(0°,10°,20°,30°)remote sensing data,the vegetation form was divided into discrete and continuous types by using the local maximum method,and the understory terrain corresponding to the vegetation form was inversed by using the point cloud distribution filter and the map forest density algorithm respectively.The semi Gaussian fitting(HAGFVC)method was used to separate the soil background and vegetation information,and the elevation difference threshold was set to extract the understory vegetation.Combined with different angle images,the patch expansion method in morphology was used to make up for the error of canopy occlusion,and then quantified the understory vegetation coverage.The results showed that for both of vegetation types,the terrain inversion results have high accuracy,the RMSE of discrete sample plot was 0.400 m and that of continuous sample plot was 0.518 m.For single angle,the estimation accuracy based on the 20°image was the highest(R^(2)=0.459,RMSE=0.166);the estimation accuracy based on 30°image was the lowest(R^(2)=0.337,RMSE=0.243).With multi angle images coupled(0°+10°+20°+30°),the estimation accuracy was significantly improved,R^(2) reached 0.675 and RMSE was 0.102.The results of this paper could provide technical support for UAV remote sensing investigation of understory vegetation information in the red soil area of South China.