Apparatus and method for searching for global minimum of point cloud registration error

A method for searching for a global minimum of a point cloud registration error, comprising:

• segmenting each of the target point cloud and the source point cloud into predetermined voxel units;
• calculating normal vectors for respective voxels;
• calculating multiple first transformation matrices for matching normal vectors of the target point cloud to respective normal vectors of the source point cloud;
• collecting multiple registration results in which a registration error between a source point cloud and a target point cloud is a local minimum as candidates;
• and
• selecting a registration result in which the registration error between the source point cloud and the target point cloud is a global minimum, among the candidates,
• wherein
• collecting the multiple registration results comprises repeatedly performing
• initializing the source point cloud and the target point cloud based on one of the multiple first transformation matrices;
• and
• registering the initialized source point cloud to the initialized target point cloud, thereby searching for a registration result in which a registration error is a local minimum.

A method for searching for a global minimum of a point cloud registration error, comprising:

• calculating multiple first transformation matrices for randomly transforming a source point cloud and a target point cloud;
• initializing the source point cloud and the target point cloud based on a first one of the multiple first transformation matrices;
• searching for a first registration result that makes a registration error between the initialized source point cloud and the initialized target point cloud converge to a local minimum;
• adding the found first registration result to candidates;
• initializing the source point cloud and the target point cloud based on a second one of the multiple first transformation matrices;
• searching for a second registration result that makes a registration error between the initialized source point cloud and the initialized target point cloud converge to a local minimum;
• adding the found second registration result to the candidates;
• initializing the source point cloud and the target point cloud based on a third one of the multiple first transformation matrices;
• searching for a third registration result that makes a registration error between the initialized source point cloud and the initialized target point cloud converge to a local minimum;
• adding the found third registration result to the candidates;
• and
• returning to initializing the source point cloud and the target point cloud based on a fourth one of the multiple first transformation matrices or selecting a registration result in which a registration error between the source point cloud and the target point cloud is a global minimum, among the candidates, depending on whether predetermined conditions for termination are satisfied,
• wherein calculating the multiple first transformation matrices includes
• segmenting each of the target point cloud and the source point cloud into predetermined voxel units;
• calculating normal vectors for respective voxels;
• and
• calculating multiple first transformation matrices for matching normal vectors of the target point cloud to respective normal vectors of the source point cloud.