3D object reassembly using region-pair-relation and balanced cluster tree
Sheng-Hui Liao1, Chao Xiong1, Shu Liu1
1School of Computer Science and Engineering, Central South University, Changsha 410083, China.
Computer Methods and Programs in Biomedicine
|September 24, 2020
Summary
This study presents a novel method for 3D object reassembly, enabling accurate reconstruction of broken objects and bone fractures. The approach achieves fast, automated, and semi-interactive reassembly, significantly improving efficiency.
Area of Science:
- Computer Vision
- Computational Geometry
- Medical Imaging
Background:
- 3D object reassembly is crucial for applications like surgical planning and restoration.
- Existing methods often lack robustness or efficiency for complex tasks.
Purpose of the Study:
- To develop a general framework for 3D object reassembly.
- To address both conventional object reconstruction and bone fracture reduction.
Main Methods:
- Introduced a rotation- and translation-invariant region-pair-relation descriptor.
- Utilized a balanced cluster tree for object segmentation.
- Developed a coarse-to-fine reassembly strategy with recursive matching quality estimation.
- Implemented least squares for precise matching and semi-interactive manipulation for complex cases.
Main Results:
- Achieved high-accuracy reassembly of broken objects within 10 seconds.
- Enabled automatic reassembly of fragments of varying sizes.
- Provided precise matching for bone fractures with cancellous structures in 30 seconds using semi-interactive operation.
Conclusions:
- The framework supports automated and semi-automated object reassembly.
- Demonstrated significant advantages for medical applications, particularly in fracture reduction.
- Offers a more efficient alternative to intensive manual processes.
