Buildings require regular maintenance, and augmented reality (AR) could advantageously be used to facilitate the process. However, such AR systems would require accurate tracking to meet the needs of engineers, and work accurately in entire buildings. Popular tracking systems based on visual features cannot easily be applied in such situations, because of the limited number of visual features indoor, and of the high degree of similarity between rooms. In this project, we propose a hybrid system combining low accuracy radio-based tracking, and high accuracy tracking using depth images. Results show tracking accuracy that would be compatible with AR applications.