Statistics for A Cross‐Dimension Annotations Method for 3D Structural Facial Landmark Extraction