This study investigates a cross-sensor point cloud registration methodology, utilizing roadmarks as robust semantic features due to their consistent visibility and uniform distribution in urban areas. The approach integrates data from ground-based mobile mapping systems (MMS) and unmanned aerial vehicles (UAVs). Key steps include selecting roadmarks, establishing graph nodes at their centroids, forming geometric structures based on node distances and angles, applying the Hungarian algorithm for bipartite graph matching, and minimizing an energy function to ensure geometric consistency. Experimental results, using data from a Velodyne VLP32 laser scanner and a DJI P4P UAV, show that the method achieves a 70% success rate in matching after extensive rotation and scaling tests. For real data, the maximum error between two sensors was reduced from 81 cm to 18 cm post-registration, demonstrating improved accuracy and reliability in multi-sensor integration.