A Ground Refinement Point (GRP) is similar to a Ground Control Point (GCP) in that it uses a GPS coordinate collected on the ground and applied to the orthomosaic to improve its accuracy. The main difference is that a GRP is applied only after the orthomosaic is produced whereas GCPs are applied before the orthomosaic is built. While GCPs require an operator to identify the target on dozens of images from different angles, GRPs require only one click per target. GRPs are faster to apply and calculate, so what is the difference in accuracy?

To determine the accuracy difference between GRPs and GCPs, two missions were flown in different conditions. Each was calculated first using only the geotags produced by the autopilot and again using GCPs collected using a survey grade RTK GPS receiver. The orthomosaics produced using only geotags were then recalculated with GRPs. The same targets and coordinates were used to apply both GCPs and GRPs.

Several ground targets are not used in either the GCP or GRP calculation but only for error checking. We measured the distance between the measured coordinates and the ground target location on each orthomosaic to determine the Root Mean Square Error (RMSE) for each mission and processing method.

  • Mission 1 was a small area with only 5 ground targets available. For the GCP and GRP calculations, 3 targets were used as corrections and all 5 are used to compare error.
  • Mission 2 had 10 targets available, 5 were used for corrections. We calculate here the error of only those targets not used in the calculation.

See the results of each test below in Tables Mission 1 and Mission 2.

Table 1: Mission 1
TargetGCP Error (m)GRP Error (m)
RMSE (m)0.03040.0680
Table 2: Mission 2
TargetGCP Error (m)GRP Error (m)
RMSE (m)0.04820.0928

The magnitude of error is about 1x GSD for GCPs and about 2x GSD for GRPs. Because GCPs are placed during processing, they become a part of the solution as the ortho-mosaic is constructed. The extra data can contribute to better image alignment and lens corrections. GRPs by contrast, can only modify an existing solution, so it’s no surprise that they are slightly less accurate. Still, comparing to the error when using geotags alone, it is an order of magnitude improvement. Because GRPs can be applied and calculated so quickly compared to GCPs, they still serve a useful function for many applications where survey-grade accuracy is desirable but not an absolute requirement.

GRPs can be applied in two ways. First, similar to GCPs, by entering the GPS coordinates manually. Second, by choosing matching visual features from another set of imagery. The other imagery can be either Google Maps imagery or another mission flown previously. This means GRPs can be applied to old missions not specifically prepared to use GCPs. Imagery can also be aligned precisely to Google Maps imagery for aesthetic purposes, or to align missions together to see changes over time more clearly.

To use GRPs, create a mission on the Drone Data Management System™ and select the GRP app from the mission’s analysis section. Learn more on the GRP knowledge base page.