Object-based 3D building change detection using point-level change indicators

•Point-level change indicators are calculated for object-based building change detection.•Slice-voxels are ultilized for building object extraction, overcoming the uneven distribution and sparsity of the point cloud.•A supervised change type determination method is proposed to detect multi-class change types.•The efficiency of the proposed method is validated both on a simulated dataset and a real airborne laser scanning (ALS) dataset. With the rapid expansion of urban areas in both horizontal and vertical directions, the complicated building structural changes challenge the existing 3D change detection methods. The existing 3D change detection methods are mainly based on local differences and rely on setting thresholds and rules, and face difficulties when determining complex change types. In this paper, to solve these problems, we present a building object extraction method using change indicators and an object-based change type determination approach. The key steps are as follows: (1) point-level change indicators are generated using the local geometric differences between the point clouds from two epochs; (2) change indicators are used to guide the process of region growing and graph cuts for building object extraction; and (3) the object-based change types are determined by a random forest classifier, relying on the elaborate features of the building objects. Experiments were carried out on a simulated dataset and a real airborne laser scanning (ALS) dataset. The proposed method achieved the best performance on the simulated dataset, and the averageprecision and recall on the real ALS dataset reached 91.1% and 85.3% respectively, which demonstrates the effectiveness of the proposed method. This work enables the 3D updating of urban building maps and can be applied to building safety monitoring and identification of potential illegal structures.