Calibrate Point Cloud

The algorithm computes the projection matrix for a dense point cloud.

Input

A dense point cloud with identity transformation matrix.

Output

Updated input point cloud.

Description

The algorithm uses the correspondences given by the point cloud points and the associated 2D pixel locations to compute the best-fit projection matrix. If useAllPoints is true, all points are used for the estimation. Otherwise only 50 evenly sampled points are chosen. The input point cloud is modified to contain the computed intrinsics and an updated transformation matrix.

The point cloud points and normals are transformed so that they are defined in the newly defined local camera coordinate system and the pose is set so that the points’ world position remains the same.