Name
Abstract | ||
|---|---|---|
3D scene reconstruction is an essential process in many computer vision fields, such as multi-robot mapping and localization. When robots capture frames in different place, which entails estimating the transformation between the sequence of frame. These issues make the 3D scene reconstruction more complication. In order to overcome such difficulty. An improved approach to 3D scene reconstruction based on fast point feature histograms (FPFH) and iterative closest point (ICP) method was proposed in this paper. First, by changing the weight calculation formula, using matched improved FPFH descriptors extracted from depth images as the initial alignment estimate can provide more robust and accurate position compared with conventional matching method. Second, the best-bin-first (BBF) is used to reduce the data dimension, which greatly accelerates the ICP iteration speed of massive point cloud data. The experiments results show that the method can quickly and effectively create complex color point cloud models of recorded scenes or objects, and has high practical value. |
Year | DOI | Venue |
|---|---|---|
2020 | 10.1016/j.micpro.2020.103064 | Microprocessors and Microsystems |
Keywords | DocType | Volume |
Point cloud matching,FPFH,BBF,ICP | Journal | 75 |
ISSN | Citations | PageRank |
0141-9331 | 0 | 0.34 |
References | Authors | |
0 | 3 | |