6 September 2017 Improving 3D registration by upsampling of sparse point cloud through fusion with high-resolution 2D image
Author Affiliations +
Abstract
This paper describes a 3D point cloud upsampling method using the fusion of 2D (e.g., EO, electro-optical) and 3D (e.g., LIDAR, light detection and ranging) sensors, that can improve the performance of 3D registration of the input point cloud with a known 3D model. In this method, a denser 3D point cloud is generated by using the corresponding EO pixel intensity values in the upsampling process. In order to increase the upsampling accuracy based on the scene complexity of a local surface area, the EO pixel entropy of the local area is used. Depending on the local entropy values (low, medium, and high), we apply different upsampling procedures (mean upsampling, full upsampling, and no upsampling respectively). By using the proposed method of upsampling, missing holes in the point cloud are filled in and the overall point density is increased, which results in improved accuracy in 3D registration of the input point cloud with its known 3D model.
Conference Presentation
© (2017) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Hyukseong Kwon, Hyukseong Kwon, Kyungnam Kim, Kyungnam Kim, Jean Dolne, Jean Dolne, } "Improving 3D registration by upsampling of sparse point cloud through fusion with high-resolution 2D image", Proc. SPIE 10410, Unconventional and Indirect Imaging, Image Reconstruction, and Wavefront Sensing 2017, 104100I (6 September 2017); doi: 10.1117/12.2276390; https://doi.org/10.1117/12.2276390
PROCEEDINGS
8 PAGES + PRESENTATION

SHARE
Back to Top