影像重叠率和地面分辨率对基于无人机影像的树高估测精度的影响

植被冠层高度是一个重要的生态度量指标；无人机遥感技术为森林和城市景观中的树高快速估测提供了更经济、高效的途径，但目前基于无人机影像的采集条件对精确获取森林结构参数的影响还存在许多不确定的因素。本文以福州大学旗山校区为研究区，采集不同地面分辨率/航向/旁向重叠率无人机影像（5cm/90%/80%、10cm/90%/80%、5cm/65%/60%），并实测了树高及冠幅。在软件中对无人机影像进行处理，生成研究区正射影像和三维点云；然后评价生成的地面数字高程模型的精度以及基于点云的局部最大值法提取树高；最后比较不同影像重叠率及地面分辨率下的树高估测精度。实验结果表明：较小的地面分辨率及较大的重叠率下的采集参数不仅生成的点云密度较高，重建更为完整，单木检测的F测度及树高估测精度更高，尤其是对于高度和冠幅较小的梧桐，而且当林下无灌木丛时有助于减小DEM误差。

Effect of overlap and ground sample distance on the accuracy of tree height estimation based on UAV images

The tree canopy height is an important vegetative structural parameter. Unmanned Aerial Vehicles (UAV) remote sensing technology is a more economical and efficient method for fast estimation of tree height in forest and urban landscape. However, there are still many uncertain factors about the impact of current UAV-images collection conditions on the accuracy of forest structure estimation. This paper selects the Qishan Campus of Fuzhou University as the study area and the images with different forward/side overlap rate (90%;80%, 65%;60%) and ground sample distance (GSD) (5 cm, 10 cm) were collected. Moreover, the field measurements for tree height and crown width were performed. First, UAV images are processed by software to generate orthophoto maps and 3D point cloud;?Second, the accuracy of the digital elevation model (DEM) was evaluated. Next, the local maximum filter (LMF) was used to extract the tree height. Final, the accuracy estimations for tree height with different GSDs and overlap rates were compared and analyzed. The results show that we can not only generate higher point cloud density and realize more robust reconstruction, but also get a higher accuracy of F measurement and tree height estimation for individual tree detection with smaller GSD and higher overlap rate, especially for F.platanifolia with lower height and less canopy. Besides, it also helps reduce DEM errors when there are no bushes under the trees.