Conventional laser scanning typically uses sensors to estimate camera pose and complete 3D reconstruction. For some cases, linear scanning is used to achieve point cloud stitching for less computational costs, however, seriously affects the practicality and flexibility. Moreover, it can only achieve semi-dense reconstruction. This paper proposes a laser scanning based 3D reconstruction combining with Simultaneous Localization and Mapping (SLAM) to tackle aforementioned challenges, SLAM provides pose estimations and accurate point cloud stitching outputs. The pipeline of the proposed method is twofold:(a) a three-dimensional semi-dense point cloud reconstruction of a target object using SLAM-based laser point cloud stitching; (b) a robust 3D-3D alignment scheme, which makes the point cloud obtained by SLAM can be merged with the laser scanning result. The results back-projected onto the original image have very limited deviations after point cloud fusion. The experiments demonstrate that our method can provide a high accuracy rate and produce a high-quality 3D surface with fine geometric details.
|Number of pages||7|
|Journal||Test Engineering and Management|
|Publication status||Published - 1 Nov 2019|
- 3D reconstruction
- equal-scale mapping
- laser scanning
- Simultaneous Localization and Mapping (SLAM)