In this paper, we propose a novel method that updates an HD map using a 3D point cloud map for the localization of autonomous vehicles. Many algorithms are needed to implement an autonomous vehicle, such as global, local planning, perception, and localization. In addition to the accurate localization system, the real-time processing is mandatory for safe driving. Therefore, the autonomous vehicle cannot use a full 3D point cloud map. Thus, most researchers for the autonomous vehicle utilize the compact HD map extracted from the 3D point cloud map. By using the HD map, the autonomous vehicle can achieve real-time processing. However, the provider of the public HD map usually only focuses on the planning. Therefore, the public HD map does not have enough information for the localization of the autonomous vehicle. As a result, we propose a novel HD map updating method that can be utilized for the map-based localization.