Light detection and ranging (LiDAR) mapping is generally utilized to obtain geotechnical and engineering data by survey in underground mines. In the traditional survey method, 3-dimensional point clouds obtained from LiDAR scanning are registered with the help of ground control points (GCP). This is a common method, and it takes time to install these points in the field and process them for registering the scans to mine survey coordinates. Recently, partial registration applications are used to combine point cloud scans. Although this process does not directly register mine tunnel scans to the mine coordinates, it is practical for combining the point clouds in the local coordinate system of the scanner. This is preferred for quick assessments that do not require precise measurements in underground operations. However, point clouds of tunnels contain a large amount of data which causes partial registration difficult. In this study, an alternative approach for registering scans accurately is introduced. For this purpose, the rock bolts on the surface of the mine tunnel were detected and used to assist the pairwise feature learning process in the proposed partial registration algorithm. The learned transformation matrix was then used to partially register the entire mine tunnel scan. As a result of this study, partial registration is applied to a tunnel scan in a more robust way with the help of detected rock bolts.