Point Cloud registration for 3D mapping

This project implements a robust point cloud registration algorithm for 3D mapping. Registration of two point clouds gives the rotation matrix and translation vector between the two point clouds. The pipeline first obtains an initial guess of the transformation matrix by matching FPFH features between two point clouds (and rejeting outlier correspondences using RANSAC) and then refining the alignment using the Iterative Closest Point algorithm.

![not-aligned] (not_aligned.png)

(Above) Three point clouds, superimposed on top of each other.

![aligned] (aligned.png)

(Above) All three point clouds aligned (registered) to the first point cloud.

For implementation details see the project report in the 'info' folder.