Control of distributed segmentation of indoor point cloud via homogenization clustering network |
| |
| Institution: | 1. School of Safety Science and Emergency Management, Wuhan University of Technology, Wuhan 430070, China;2. School of Artificial Intelligence, Wuchang University of Technology, Wuhan 430223, China;3. Changjiang River Science Research Institute, Wuhan 430010, China;4. Chongqing Academy of Metrology and Quality Inspection, Chongqing 401121, China |
| |
| Abstract: | Indoor point cloud segmentation is necessary for many applications including object identification for indoor navigation, facility management and so on. In order to construct accurate segmentation system of indoor point cloud, this paper improves cloth simulation and uses it to control the segmentation of ceiling and floor. Simultaneously, we obtain the wall points according to the points’ density. We mainly introduce indoor objects control segmentation via clustering strategy. First, we generate the cutoff distance according to the angular resolution and scan distance. Second, we use exponential function to determine the local density. Third, the cluster center is determined according to the magnitude of product of local density and distance which are normalized. Finally, the points affiliated to a cluster are controlled one by one according to the distance between point and cluster center. Segmentation of objects is then realized based on the established clusters. Experiments indicate that the proposed algorithm achieves a competitive performance when compared with several state-of-the-art algorithms. The performance of the proposed method and the accuracy of distributed segmentation are affected by the degree of closeness between objects. |
| |
| Keywords: | |
| 本文献已被 ScienceDirect 等数据库收录！ |
|
