ETRI-Knowledge Sharing Plaform

KOREAN
논문 검색
Type SCI
Year ~ Keyword

Detail

Conference Paper 3D Object Classification and Segmentation from Large Scale Area Reconstruction
Cited 0 time in scopus Share share facebook twitter linkedin kakaostory
Authors
Hye Sun Kim, Yun Ji Ban, Chang Joon Park, Ho Won Kim
Issue Date
2023-10
Citation
International Conference on Information and Communication Technology Convergence (ICTC) 2023, pp.1-3
Publisher
IEEE
Language
English
Type
Conference Paper
DOI
https://dx.doi.org/10.1109/ICTC58733.2023.10393235
Abstract
3D reconstruction is difficult to use in general applications because it treats objects as a whole without distinction. Post-processing to segment them into individual objects is essential. However, when the reconstruction target is a large-scale area, it is difficult to segment directly from the 3D geometry because it is divided into small rectangular tiles for convenience. We propose a method that can accurately classify and segment objects even in the case of locally divided 3D tiled geometry. Instead of using 3D segmentation, we adopt the method of 2D instance segmentation of a global multi-view image and then projecting it onto the 3D geometry. This solves the problem that objects located on the boundaries of tiles are fragmented and not properly segmented.
KSP Keywords
3D Reconstruction, 3D Segmentation, 3D geometry, 3D object, Object classification, Post-Processing, large-scale, multi-view images