Research Article

A Dual Discriminator Adversarial Learning Approach for Dental Occlusal Surface Reconstruction

Algorithm 1

Standardized processing of the dental model posture.
 Input: P0, P1, and P2 are the three feature points on the first molar, the second molar, and the second premolar, respectively; N0 is the normal vector of the point P0.
 Output: tooth model posture (x, y, z).
 Step 1: construction of local coordinate system: the normal vector N0 of P0 is calculated, and then, the local coordinate system of the tooth model is constructed by P1, P2, and N0.
 Step 2: adjusting the normal vector of the occlusion surface: firstly, the translation matrix T  from the origin P0 of the local coordinate system (P1, P2, N0) to the world coordinate system (X, Y, Z) is calculated, and then, the rotation matrix R between the z-axis of the system (P1, P2, N0) and system (X, Y, Z) is calculated. Finally, the compound matrix RT is applied to the tooth model to adjust the position of the occlusion surface.
 Step 3: automatic axis setting by bounding box: firstly, the central point o of the bounding box of the tooth model is calculated, and the local coordinate system (x, y, z) with point o as the origin is constructed. Then, point o is moved to the origin O of system (X, Y, Z); finally, the long axis y-axis of the bounding box is parallel to the Y-axis of system (X, Y, Z) by the Euler angle transform, so as to complete the standardized processing of the dental model posture.
 End.