I have 3D object (soft surface like aircraft wing) and want to use the deep learning tools to register the object using ICP, perform 6 Dof translation and rotation on the object. I have studied many libraries, methods of pose estimation, filtering, normalization, and image rendering but still not clear what are the steps or general pipeline? like someone using Intrinsic shape signatures ISS for capture the points of the object, or (SHOT) feature description and Random Sample Consensus (RANSAC) feature matching method to coarse registration.
My tools are: IMU, depth camera (D435), laser for tracking and HoloLens 2.
I need help to understand what the general required steps are, based on what I decide to select algorithm.
Many thanks in advance.