Algorithm 1: The Procedure of the Proposed Pose Tracking Method

Input: The model point cloud $P_m$, the current k-th sensor point cloud $P_s^k$ While $k>1$ 1: Convert the $P_s^k$ to the depth map $I$. 2: Detect the lines collection $L$ in the depth map $I$. 3: Calculate the lines descriptors for each line. 4: Sort the lines by the line length $len$ in descending order. 5: Select the first $Q_c$ lines and calculate the line similarity degree matrix $SDM$. 6: Calculate the roll angle variation $\Delta \varphi$ and the transformation matrix $H_{roll}$. 7: Perform the point cloud simplification process for $P_s^k$ and obtain the sparse sensor point cloud. 8: Calculate the transform matrix $H_{icp}$ by aligning the sparse sensor point cloud with the $P_m$. 9: Calculate the current k-th transformation matrix $H_k$. 10: Calculate the six-DOF relative pose parameters. 11: $k=k+1$, and go to step 1. end Output: the six-DOF relative pose, including the roll angle $\phi$, the pitch angle $\theta$, the yaw angle $\psi$, the $\Delta x$, the $\Delta y$, the $\Delta z$.