Deformable tree matching with tangent-enhanced coherent point drift

摘要

Deformable, anatomical trees represented by scan data from different times are matched. Coherent point drift (CPD) solved using expectation maximization is enhanced with tangent or other curve information. By including point-curve information, another characteristic than GMM-based probabilities are included in the cost function for matching. The angle information provided by the tangent, normal, or other point-curve measure may more likely match points in one set representing a tree to points in another set representing the tree.