Analysis and benchmarking of gravitational search algorithm against iterative closest point algorithm for point cloud registration

Having two cloud points P and Q, the point cloud registration aims to find a transformation (rotation and translation) to apply to P to match Q as good as possible. The iterative close point is a well-known algorithm in solving the point cloud registration problem. This iterative algorithm computes the mean, correspondence value, cross-covariance matrix, and single-valued decomposition to find the rotation and translation. This approach also can be easily adapted with an optimisation algorithm such as gravitational search algorithm. To solve the point cloud registration using gravitational search algorithm, error function such as mean square error can be used as objective function. If the cloud points have outliers, the objective function could be modified with a kernel function to deal with those outliers. During the first half of this project, the iterative close point algorithm is implemented as a benchmark to solve point cloud registration based on different cloud points of various level of difficulties. After that the cloud point registration is solved based on gravitational search algorithm. Finally, a statistical analysis is performed to analyse the performance of the two approaches.