We are running a pilot project to prove that a model produced using aerial photography and Pix4dmapper can provide results that are within a centimetre of the results produced by a Total Station. To do this we would like our point cloud and mesh to use the location of the Total Station as the datum for the coordinate system.
We are using GCPs from Google Earth. Ideally we would like to use WGS84 for our images and GCPs and an arbitrary system for our point cloud and mesh, but other approaches that will work are okay.
We have been able to produce a point cloud in an Arbitrary output coordinate system from which we can confirm that distances in our model match those from the Total Station within an acceptable margin of error.
Where we need help is changing the origin and orientation of the output coordinate system to match the Total Station. We have tried using Site Calibration, but this does not seem to produce the desired results – or we are not using the correct approach to selecting the values we enter.
So far we have been creating an MTP at the location of the Total Station and another at the back-sight and using the location of these MTPs and the orientation of a polyline between them to determine the factors to enter the Site Calibration parameters. However, the resulting coordinates are incorrect and the point cloud and GCPs are sometimes completely separated.
Any advice or guidance would be greatly appreciated.
Hello Ian,
I am currently working on something similar to also proof aerial photography for surveying. Maybe we can work together on this or maybe I can help you out. Feel free to email me at nkariger@gmail.com
Nick.
If you want to use an arbitrary local system for the Total Station, why not just use the point cloud to define the Total Station coordinates, the Back-site point coordinates, and a few additional points ? Then use those coordinates as the data file for your Total Station. Then future side shots from the Total Station should match (within the accuracy and precision of your model) the x,y,z of a MTP at the side shot location in the model. It’s the GCP concept in “reverse”.
If this does not produce realistic results, then the Z plane of your Model is likely incorrect due to in-precise GCP’s from Google Earth. Any model rotation about the Z-axis would show up as error when positions are compared to positions shot from the total station.
Another thought: You could re-process the model and replace your Google Earth GCP’s with the point file from the total station? This would move the model to the local arbitrary system of the total station points.
If you have access to a survey grade GPS, your pilot trials would be much easier. Your 3D model would be in real world coordinates (I much prefer a Grid State Plane System) and every point in the Model is reproducible with a Total Station or RTK GPS.
I will assure you that Pix4D (and a couple others) can produce a point cloud with sub-centimeter accuracy, I’ve done it many times. I use 5 GCP’s (multiple observations w/ RTK GPS) for a typical site plan, fly low altitude with very high overlap, and add many MTP’s to the Model. Pix4D is the most user-friendly solution in my opinion.
Ryan Fontaine why do you need to add MTPs to the model?