DroneMapper Labs: Geospatial Accuracy – 8/24/16
Here we report on a set of recent imagery collections utilizing primarily UAV platforms (one manned platform included – Project L) with varied collection sensors. In each case well surveyed ground control was used in the processing of the DEM and orthomosaic. In each of the DEM/ortho sets the position of the control was compared to the surveyed coordinates to compute horizontal and vertical root mean square errors (RMSEs). All measurements were accomplished utilizing Global Mapper, version 16.2. The table below illustrates seven data sets with RMSEs computed.
Examples of DroneMapper Ortho & DM Ground Control Geo-spatial Errors (RMSE)
The data sets include a manned aircraft using Vexcel’s UltraCam Falcon prime (Project L), two fixed wing UAVs using Sony cameras (Projects B and WE) and four DJI quad-copter UAVs using the FC series of sensors (Projects CB, TB-1, TB-2 and McB). In all cases except one, Project B – vertical error, the horizontal and vertical RMSEs are sub-pixel (compared to the native GSD).
When we average the four DJI project horizontal and vertical errors as a percentage of GSD we see that one could expect an error of one-half pixel horizontal and three-quarters pixel vertical. In terms of absolute map accuracy with 95% confidence one could expect 1.2 pixels horizontal and 1.5 pixels vertical. So for you DJI operators, if your application requires absolute vertical accuracy of 3”, plan on imaging at 2” pixel GSD or less with well surveyed control.
More information on the GCP Reporting tool here: https://dronemapper.freshdesk.com/solution/articles/2100030976-generate-a-gcp-report-with-elevation-rmse
Marked as spam