Point Cloud Registration for 3D Modelling

In 3D imaging, there are instances where only a particular view is visible to the point cloud acquisition device. One can get only a certain point of view from the fixed camera system. However, we can obtain multiple such views from a fixed camera system from different perspectives. This paper deals with a system that has improvised algorithms consolidated together to get a properly aligned model of these different perspectives. This system is designed to combine different point clouds together via matching the points and make a whole complete model in one point cloud dataset. The proposed solution delves into affine transformations and Iterative Closest Point (ICP) Algorithms. Affine transformation preserves collinearity. The problem of consistent aligning of 3D point data is known as the registration task which is resolved using ICP. Also, the Nearest neighbors algorithm is used. The complete project was idealized, designed, and demonstrated in python with minimal requirements. For visualization, the transformed arrays are extracted into the 3D output files which can be viewed on in- house or third-party rendering software.