| Autor: |
Wang, Chen, Xu, Yuhua, Wang, Lin, Li, Chunming |
| Předmět: |
|
| Zdroj: |
Visual Computer; Dec2022, Vol. 38 Issue 12, p4279-4290, 12p |
| Abstrakt: |
The indoor scenes usually exhibit strong structural regularity; on the other hand, the local geometric features are scarce and difficult to be distinguished. The properties make the solutions of the existing global registration algorithms ambiguous. In addition, finding global optimal solution is usually computationally expensive. These problems make the global registration of indoor point clouds a challenging task. Based on the Manhattan-world assumption and color information, we propose a fast and effective algorithm for the global registration of indoor colored point clouds in this paper. We introduce the Manhattan-world assumption into the global registration of indoor point cloud to limit the number of all possible rotation solutions. As a result, our algorithm solves the problem very fast. Furthermore, we utilize the color information hierarchically to eliminate the ambiguity induced by the structural models. The limited rotation set is filtered by the global color histograms, and the final solution is evaluated by the local features. The experiments demonstrate that our algorithm can complete the global registration for the point clouds with hundreds of thousands of points in 0.1 second with comparable accuracy with others. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
Full text from SpringerLink