Settings for orthomosaic processing from GoPro

Hello,

I am usually processing orthomosaics for further displacement/volume change analysis from UAV data with Pix4D mapper. I´ve now been trying to do the same from GoPro videos. I have several videos from a flight with a plane that had 4-5 GoPro cameras mounted to it, looking in different directions for different angles of the same area. I exported 2 frames per second (with EXIFtool) from all those videos and now have a few question on how to further process them in Pix4D, to get the best orthomosaic I can regarding georeferencing (and resolution):

  • Since its multiple sets of images from 4-5 cameras, do I still load them all in one project at once or does it make more sense to process them singularily and then combine them later on?
  • For Processing Options in 3D Maps, is there anything I should consider specifically, since the images are frames from a Vvideo taken from a plane that was not always flying in a stable height/angle…?
  • I don´t really have GCPs from the area, only a set of estimated ones, mapped from a orthomosaic provided by the province. Is that still better then using no GCPs, or will it make things worse regading georeferencing
  • My Point clouds tend to have a lot of noise due to shadows and steep terrain (flight was over a high mountain area) and because the plane turning while flying I have quite a lot of sky in my point clouds. I haven´t found anything on how to filter the point clouds in a systematic way, except for doing it manually which turns out to be quite difficult in non flat terrain.

A lot of questions… any thoughts and ideas on how to improve my workflow would be really appreciated!

@anne.hartig You can process the data obtained from the GoPro; however, using drone imagery would be a much better option compared to GoPro images.

If you still prefer to process the GoPro data and evaluate the results, I recommend reviewing the following support article:

When selecting images, choose the camera that is pointing directly toward the ground rather than those captured at larger oblique angles. Only combine data from two GoPros if there is insufficient overlap.

You can use the 3D Maps template for processing. It is also better to use GCPs extracted from an orthomosaic generated with GCPs, as this will improve accuracy.

Based on the flight pattern and conditions you described, some point cloud noise is expected. You may need to spend significant time cleaning the point cloud to achieve usable results. For this reason, I recommend using a drone for this project, as it would provide better quality data and reduce post-processing effort.