基于无人3D摄影技术的雪松(Cedrus deodara)群落高度测定

植物群落高度是反映植物群落特征的重要指标,植物群落高度的测定能给植物群落多样性分析、生物量估算、功能形状研究提供重要的数据基础。传统的森林调查主要由生态调查工作者通过目测或者利用激光测高仪对每个个体进行逐一测定,因此劳动强度大,耗时费力,并且难以进行大面积的植物群落高度调查。近年来,随着无人机(Unmanned Aerial Vehicle,UAV)技术的飞速发展,催生了无人机低空摄影测量和遥感技术,该技术已被应用于农作物植株高度测定和生物量估测等。然而针对植被类型多样、树木种类繁多且地形复杂的山区如何精确的获取植物群落高度仍然是一个较大挑战。以缓坡地形的云南大学呈贡校区为研究区,选取校区内人工种植的雪松(Cedrus deodara(Roxb.)G.Don)林为研究对象,利用无人机搭载可见光相机平台获取研究区近地面航空影像,利用高分辨的影像匹配加密获得的点云数据生成数字表面模型(Digital Surface Model,DSM)。依据点云分类提取非植物类点,消除少数因植被与非植被相接的边缘模糊而错分类的部分,内插后生成数字地面模型(Digital Terrain Model,DTM)。将DSM和DTM叠加相减得到树木高度变化模型(Canopy Height Model,CHM),即获得研究区各个雪松的高度。然后利用激光测距仪测定研究样地内100棵雪松的高度,将此测定的树高与无人机航测技术生成的CHM模型测定的树高值进行精度检验。结果表明无人机测定的树高值与激光测距仪测定的树高值线性拟合度较高,r2值在0.904以上。此方法基于无人机影像生成空间模型,提取树高,受外界环境因素影响较小,且成本较传统测树方法低廉,可广泛运用于各种植物群落的调查研究当中,有极好的应用前景。

Height measurement of a Cedar(Cedrus deodara) community based on unmanned aerial vehicles(UAV) 3D photogrammetry technology

The height of plant communities is an important parameter, which can be used to estimate biomass and is a functional trait of plant communities. Tree height and plant community height are commonly estimated by forest surveyors or measured through the use of a laser altimeter in traditional forest surveys. However, this traditional survey method is hard to apply to measure tree and plant community height in large areas. Recently, with the rapid development of unmanned aerial vehicle(UAV) technology, the development of low-altitude unmanned aerial photogrammetry and remote sensing technology has been stimulated. Previous studies have shown that UAV images can used to measure crops and orchard height, and to estimate biomass. However, obtaining the height of vegetation in rough mountain areas remains a big challenge. Chenggong campus of Yunnan University was chosen as an experimental area because of the small hill area. We also selected a plot in which 4203 m² was dominated by pine trees(Cedrus deodaraRoxb.] G. Don). Aerial images of the experimental area were acquired using the UAV equipped with a common digital camera platform(Sony ILCE-7R). We aligned these images to obtain point cloud data and to produce dense point cloud data. Then we built a digital surface model(DSM) using these point data. We extracted non-plant points according to the classification of cloud point. Some sections were removed because of edge fogs between the plant and non-plant parts, and a new height model(digital terrain model; DTM) was built by interpolation. For the height variation model of the cedar canopy in the study area(canopy height model; CHM), the height was obtained by overlaying the DSM and DTM. DTM was subtracted from DSM in order to obtain the height variation mode of the cedar canopy in the study area. The height variation model was the height of cedar individuals. Afterward, an accuracy assessment has been carried out using linear regression analysis. The heights of 100 cedar individuals, measured by a laser rangefinder were used as validation data. There are very greatly correlations(r² > 0.904) between the tree heights measured by laser range finder and quantified by CHM derived by overlaying DSM and DTM. Both the space model, which was based on the UAV images, and tree height which was subtracted from the space model were less affected by external environment factors. Additionally, this method is easy to be performed and can be widely used to investigate various plant communities and has prospects for use in ecological application.