期刊
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
卷 149, 期 -, 页码 105-118出版社
ELSEVIER
DOI: 10.1016/j.isprsjprs.2019.01.020
关键词
Hovermap; UAV; LiDAR; Drone; Mapping
类别
资金
- Australian RTP scholarship
- Australian CSIRO Data61 group
Utilised globally across a wide range of applications, the ability to assess and understand LiDAR system capabilities represents an essential component in developing informed decisions on instrument selection and the logistical planning processes associated with site-specific limitations, project objectives and UAV operations. This study employed the new SLAM-based CSIRO Hovermap LiDAR system within a purpose-built environment as a testbed to experimentally investigate the interactive effects of fundamental UAV flight parameters on key metrics of LiDAR point clouds. Assessed within a full factorial design at both Site- and Target-levels, the UAV input variables of Pattern, ground Speed and above ground Altitude (AGL) were tested against the point cloud response variables Density, GSD and Accuracy as measured by RMSE and cloud-to-mesh Euclidian distance ('Deviation'). A novel approach is described wherein the trajectory file of each flight was examined to determine the observed values of the input and response variables, remove noise and facilitate a standardised basis of comparison. Several new terms are introduced including Sampling Effort Variable (SEV, s.m(-2)), Effective Scan Rate (ESR, pts.s(-1)) and Effective Density Rate (EDR, pts.m(-2).s(-1)) as well as an alternate approach to describe Pattern (s.m(-1)) as a scalar quantity. Reporting significant effects with all response variables at both Site- and Target-levels, the Range of the LiDAR sensor, closely associated with Altitude, presented as the single most important factor. Interestingly, the combination of the independent variables as SEV and EDRpred ('predicted' EDR) showed the highest coefficient of determination in the Site-level prediction of Density (R-Adj(2) = 0.894) and GSD (R-Adj(2) = 0.978,), respectively, whilst Range best correlated with observed RMSE (R-Adj(2) = 0.948) and Deviation (R-Adj(2) = 0.963). Predictive models returned mixed results when evaluated at the Target-level and highlights the need for further investigation to achieve the maximum benefit of high-resolution UAV LiDAR. The results presented here confirm that the CSIRO Hovermap performance is robust and, although variable depending on UAV flight parameters, is predictable and demonstrates the potential value in understanding system performance in harmonised flight planning to achieve project-specific objectives.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据