I fly using an RTK drone on dirt sites of various sizes and changes in elevation. One site in particular is over 100 acres with elevation changes of about 50 feet. I set about 15-20 check points and sometimes I will have big variances in elevation with one target being +.15 (ft) and then the next being -.15(ft). I have been able to adjust this a little bit by using about 5 of the check points as 3dgcps but I’m not sure if that is giving me an accurate representation of the data or I am just forcing it to say what I want it to.
Please note that the models generated in Pix4Dmapper have two types of accuracy:
Relative accuracy : quantified by comparing individual features on the 3D model and the same features measured in the field (reality). For example, if a 3D model of a building shows windows two meters apart, and this is the same distance as on the actual building, the model has high relative accuracy.
Absolute accuracy : defined as the difference between the location of features on the 3D model and their true position on the Earth. For instance, if the same 3D model of the building is not in the correct place on the map, it has low absolute accuracy, even if it has high relative accuracy.
On one hand, the absolute accuracy of the outputs should be expected to be 1-2 pixels (GSD) horizontally (X and Y coordinates) and 1-3 pixels (GSD) vertically (Z coordinate). For example, if your project has a GSD of 3 cm, then the absolute accuracy will lie within 9 cm. To improve the absolute accuracy, you need to have an image dataset with highly accurate geotags (RTK/PPK). If your dataset has been acquired with a standard GPS, then make sure that the GCPs are properly measured, marked in as many images as possible and at a high zoom level. It is also essential to distribute them evenly across the dataset, in order to avoid any georeferencing bias.
On the other hand, the relative accuracy of the outputs in general and without GCPs should be expected to be 1-3 pixels (GSD). This can be improved locally (especially in areas with lower overlap or difficult image content) with the help of Manual Tie Points (MTPs). An MTP is a 3D point corresponding to a keypoint that is marked (clicked) by the user in the images. They are used for assessing and improving the relative accuracy.
It would be helpful if you shared the quality report with us, in order to provide more specific tips.
Thanks for the reply. As far as the GSD goes, I don’t really know how to accurately calculate that for different jobs. I have it set at 1.1 inches in Emotion but in my quality reports it always says something else so that has me a little confused.
Here is a quality report from one of my flights using targets as checkpoints only.
CARVANA 4-27-20_report.pdf (1.7 MB)
Here is the same flight using mtps and 2d GCPS
CARVANA 4-27-20 test second try_report.pdf (1.7 MB)
The GSD of your resulting model cannot be adjusted while processing your Pix4Dmapper project. The GSD value will depend on several factors, such as flight height, focal length, camera model etc. We provide a tool that helps the user calculate the GSD in advance: GSD calculator.
Hopefully this was useful.