I have collected images using gopro 5 in wide mode with good overlap and closer to the object without GPS co-ordinates in Indoor and later i have added random GPS coordinates from CSV, I cannot get the good 3d model. So pix4d computes tie point from nearby images based on GPS alignment or it does based analyzing on the texture of the images.
Pix4D generates keypoints based on image texture. These points are later on matched and become 3D automatic tie points.
if geotags exist, they will help the reconstruction and the final model will be georeferenced accordingly.
If geotags do not exist, the final model will be shown in an arbitrary coordinate system.
In this case, we would recommend to run the process without adding any external information. The result will be a 3D model with arbitrary coordinates, with no scale and no orientation. you could afterwards add an scale constraint to be able to measure real distances in the 3D model.
If the process fails again, I would add and measure some manual tie points (MTP) and/or ground control points (GCP) and try again. The GCPs and MTPs will help to find more keypoints and matches and therefor they help the 3D reconstruction.
It may occur that the process fails again, even with the help of MTPs and GCPs. In that case we would recommend to check the image overlap and in case it is not enough, going back to the site and take some more pictures.
You need more overlap…indoor work is especially difficult. Unless you are in a big open warehouse type building, the overlap rules change in tight spaces. You likely need to double or triple your picture count when indoors…