Greetings.
I just completed a mapping flight using a quadcopter and my final ortho images and the DEM does not align with my GCPs very well at all.
I ended up with just under 900 images covering an area of about 90 hectares. Prior to doing the flights, I put in 13 Ground Control points using a very nice Trimble GPS unit. I post processed the GPS data and the software tells me that all of the coordinates are good to within 5-15 cm. I have processed the imagery in Agisoft following the tutorial for v1.2 with GCPs. All of my images are geotagged but the accuracy of these geotags is questionable, but most of the images aligned. Following placement of the markers in Agisoft, I import the coordinates for each of the markers (longitude, latitude and altitude). Agisoft then calculates an "accuracy (m)", "Error (m)" and "Error (pix)." I'm not quite sure how to interpret these values or how they are derived. The error in meters is quite large, ranging from 4 to 164 meters!!?? (see attached). Given the info that I obtained when processing the GPS data, there is no way that these coordinates can be off by this much.
Can anyone explain what these accuracy and error values mean and how they are derived.
Ignoring this issue and moving on, I completed the processing to create an ortho and DEM. Within Agisoft, I've examined the results and, visually the imagery seems to line up very nicely with the GCPs (see attached example for GCP5). I'm interested in not just the agreement in the X and Y but also the alignment in the Z axis. I'm not sure if there is a way to compare this in Agisoft, so I exported both as TIFFs and brought them in to ArcGIS. I also imported the shapefile of the GCPs that the Trimble Pathfinder software creates.
In Arc, the GCPs do not line up well at all (see attached example for GCP5; green diamond is GCP location). They seem to be off by 2-9 meters or so in the X,Y direction. In Arc, I can also use the identify tool in Arc to get the X and Y coordinates for both the GCP point and the flagged location in the ortho that Agisoft creates. The coordinates for the GCP point agree with the GPS coordinates. The coordinates for the flagged location in the ortho do NOT match the GPS coordinates for this point.
I can also use the identify tool to get the height for both the GCP (recorded using the GPS unit) and the height from the DEM generated by Agisoft. They are not even close. In the case of GCP5, the GPS elevation is 157.15 meters and the elevation from the DEM for that same spot is 167.77 meters. The elevation for the flagged location of the GCP is 167.83. I'm seeing vertical errors of 2.5-10 m at the other GCPs as well.
I'm not really expecting centimeter-level accuracy here but being off by 2-9 meters in the XY and 10 meters in the Z direction seems very very bad. I need to be able to do much better than this.
Any suggestions?
Topic: GCPs and image registration (Read 15050 times)
