I use Meshroom (based on the framework AliceVision) to reconstruct a scene based on photographs (e.g. a bottle on a table in a room). The software captures the bottle, but also the table and the whole room, so I have a huge model in the end. I also a have a ground truth model of the bottle.
My goal is to align the whole reconstructed surface to the ground truth model so that the coordinate system and the scale will be known. I managed to do it "manually": Meshlab has an alignment feature which uses ICP, but I need to provide input to the algorithm (clicking on arbitrary points on the reconstructed surface and clicking on the same points on the ground truth model, like here). However, I'd like to automate the process. One of the biggest challenges is that the reconstructed model contains many details which are irrelevant to the alignment (like the table, walls of the room, etc.).
Can you suggest a way to perform this alignment without any user input?