Automatic building exterior mapping using multilayer feature graphs

摘要: We develop algorithms that can assist robot to perform building exterior mapping, which is important for energy retrofitting. In this task, a needs identify facades in its localization and mapping process, turn be used navigation. Existing rely on low level features such as point clouds line segments cannot directly applied our task. attack problem by employing multiple layer feature graph (MFG), contains five different ranging from raw key points planes vanishing 3D, an extended Kalman filter (EKF) framework. analyze how errors are generated propagated the MFG construction then apply data observations EKF map facades. have implemented tested MFG-EKF method at three sites. Experimental results show successfully constructed modern urban environments with mean relative of plane depth less than 4.66%.