Abstract: The structural stability of highway tunnels directly affects traffic safety. To monitor tunnel deformation in real time, eliminate potential safety hazards, and reduce the risk of accidents, this paper proposes a method for monitoring tunnel crown deformation based on vehicle-mounted laser point clouds. The method involves extracting the tunnel central axis and continuous cross-sections, improving point-to-plane nearest-point registration and precision evaluation techniques for accurate deformation analysis. The main steps are as follows: (1) the tunnel central axis is extracted using multi-view projection and median method, followed by fitting normal planes based on the local similarity of central axis points to clip tunnel cross-sections; (2) the tunnel cross-sections are partitioned and registered with normal and radial constraints, and the registration result is computed iteratively by weighting the registration accuracy of each partition. Experimental results show that this method improves registration accuracy by 19% and 79% over ICP and K4PCS algorithms respectively under complex scenarios. It effectively reflects highway tunnel deformation and provides important technical support for tunnel operation safety.