ABSTRACT The development of unmanned aerial vehicles (UAVs) along with that of positioning and imaging sensors has promoted the use of photogrammetric techniques for applications other than the conventional acquisition of the Earth’s surface information. Moreover, the accuracy of photogrammetric products depends on several parameters, such as sensor quality and stability, image scale, strip overlap, and the number and distribution of ground control points (GCPs). Although technological developments have improved the automation and accuracy of the photogrammetric process, the acquisition of GCPs remains a highly time-consuming task. To address this problem, this study investigated the positional error of digital surface models (DSMs) generated via the post-processing of data acquired with a SenseFly eBee Classic, with respect to the number and distribution of GCPs. A single flight was conducted over a test field located at an urban expansion region of Presidente Prudente, State of São Paulo, Brazil. Nine DSMs were generated using different configurations of the number and distribution of GCPs. An accuracy assessment was performed based on 20 horizontal check points and 270 vertical check points determined by global navigation satellite system (GNSS) positioning. The vertical error surfaces were generated by Kriging. The Student’s t-test, Chi-square test, and root mean square error (RMSE) were used for the positional accuracy assessment. The results show that the qualities of the products generated by GNSS-supported Aerial Triangulation (GNSS-AT) without GCPs were in accordance with those indicated by the UAV manufacturer. In contrast, the results obtained by considering the GNSS-AT with different GCPs were not in accordance with the manufacturer’s claims. Finally, RMSE values of approximately 3 ground sample distance (GSD) for the horizontal component and 5 GSD for the vertical component are achievable with at least three well-distributed GCPs; however, the use of additional points is encouraged.

中文翻译：

---

基于无人机的数字地面模型评估及地面控制点数量和分布的影响

摘要 无人驾驶飞行器 (UAV) 的发展以及定位和成像传感器的发展促进了摄影测量技术在常规获取地球表面信息之外的其他应用中的使用。此外，摄影测量产品的精度取决于几个参数，例如传感器质量和稳定性、图像比例、条带重叠以及地面控制点 (GCP) 的数量和分布。尽管技术发展提高了摄影测量过程的自动化和准确性，但 GCP 的获取仍然是一项非常耗时的任务。为了解决这个问题，本研究调查了通过对 SenseFly eBee Classic 采集的数据进行后处理而产生的数字表面模型 (DSM) 的位置误差，关于 GCP 的数量和分布。单次飞行在位于巴西圣保罗州普鲁登特总统城市扩展区的试验场上空进行。使用 GCP 数量和分布的不同配置生成了九个 DSM。基于全球导航卫星系统 (GNSS) 定位确定的 20 个水平检查点和 270 个垂直检查点进行了准确性评估。垂直误差面由克里金法生成。学生 t 检验、卡方检验和均方根误差 (RMSE) 用于位置精度评估。结果表明，在没有 GCP 的情况下，GNSS 支持的空中三角测量 (GNSS-AT) 生成的产品质量符合无人机制造商的指示。相比之下，通过考虑具有不同 GCP 的 GNSS-AT 获得的结果与制造商的声明不符。最后，使用至少三个分布良好的 GCP 可以实现水平分量的大约 3 地面采样距离 (GSD) 和垂直分量的 5 GSD 的 RMSE 值；但是，鼓励使用附加点。