Register (merge) components from different sources (LIDAR,Photo, OBJ)

Development Asset Creation
,
1

Well, so I am wondering if there is any possibility beside control points to merge different components into each other. I am trying to do the following:

 

Load a laserscan pointcloud (with internal registration)

Add some drone photoshots (with GPS registration)

Add a solid textured object (OBJ) generated from an LIDAR-Scan app, however with no registration as it is a simple OBJ file.

 

Merge them all together into one solid object within RealityCapture for use in Unreal Engine.

 

What I do get for now is: A totally offset OBJ within 2 other components which also do not register together. 

What is the workflow for this? How can I align (register) the OBJ into the other pointclouds?  How can I align them all together? Is the only way to use control points? How do I properly use control points with an OBJ?

I have watched several tutorials but simply cannot figure out how to manually move around the imported OBJ within the other data on the grid. I also could not figure out how to generally position components within the “map”. (So meanwhile I guess there simply is no way for this)

What I am missing is a kind of a manual registration process or a pre-registration process where I point out a close match for one component to the other to the software and it then completes the registration once I gave it a close match hint for two components. This is typical within LIDAR-software where you roughly match the scans by hand and then it´s easy for the software to find all the connecting points. If I want to do so in RealityCapture with a randomly positioned imported OBJ there is simply no way I could find to make this work. If I am using the control points instead, the software only finds completely wrong matches because the basic positioning of the objects is totally out of space.

So what can I do?

Many thanks for a hint!

 

In the picture you can see the OBJ file (interiour of a house) in purple together with the photo-pointcloud. However it does not show the LIDAR-Pointcloud, which is also not correctly registered to the photo-pointcloud until now. 

 

2

Hi 8K Vision,

you don’t merge OBJ using this way. You have to have created it in RealityCapture or you need to have the same coordinate system.

If you are scanning buildings with interior and exterior, you need to have some overlap between these parts of a building.

Otherwise, you can use ground control points with coordinates in one coordinate system for exterior and interior.

If you don’t have overlap or ground control points it will be very hard to merge these datasets.

There are some tutorials with this problem: https://www.youtube.com/watch?v=HbdXMc-NtZQ&t=321s, https://www.youtube.com/watch?v=-CwdugODkmQ&t=23s, https://www.youtube.com/watch?v=ZZut6f17Vtc&t=28s